I'm eavesdropping on a fascinating email exchange between Chris Dent (of Blue Oxen and Socialtext fame) now working on TiddlyWeb and Frank McIngvale who's project is WikklyText.

I haven't had time to take it all in yet. But it looks like Chris has got TiddlyWeb working on the Google App. Engine. And Frank has got a stand-alone TiddlyWiki markup language working.

The main idea of TiddlyWeb continues Chris's focus (since Blue Oxen days with EEKim, I'd guess) on sub-page level elements on wiki. Remember Blue Oxen's thing was Purple Numbers, individual paragraph Ids. Here, he's using "Tiddlers", the individually named, sub-page elements that TiddlyWiki would show or hide, and assembling them in a new, looser collections called "bags".

From what I understand so far, having named tiddlers rather than arbitrary purple numbers is definitely a move in the right direction. (In the sense that it makes the small pieces human-addressable as well as machine-addressable.) In fact each item is addressed by a combination of Tiddler name + bag name (where bag is more a kind of policy or query)

There's long discussion going on right now about URIs (which seem to become almost queries or operations on the bags) to access the tiddlers in a ReSTful way that I'm still absorbing.

Anyway, they're definitely "banging the rocks together" in wiki and breaking pages up into a finer granularity. And, after Twitter's discovery of the virtues of 140 character status updates, now generalized to a theory of micro-blogging, the world is definitely ready for a wiki micro-chunking experiment. Who knows where it will lead?

I use Subversion source-control (I have one web-hosted repository, one on my pendrive, and I just started using Google Code.)

But I'm tempted by distibuted source-management. The arbitrary decision as to whether I host a project online or on my pen-drive is ... well ... arbitrary, and sometimes needs to be revised. Distributed would be better. And I'm increasingly tempted by Bazaar (bzr).

The other thing that attracts : it's in Python and easily incorporated into an existing python program. I haven't given up on the whole SdiDesk / wiki-like notebook / developing in wiki thing. Perhaps there'll be a bazaar-as-PageStore one of these days.

Meanwhile ... watching the proliferation of repositories (I'd like to check out Folknology's GitHub hosted Reactores) I'm starting to think there's a need for an access-standard. Some kind of lowest common denominator ways of checking out checking in and merging codebases from all these repositories.

Anyone know of anything like this?

Ouch! Major down for Mind Traffic Control today.

Must be something to do with the fix I posted this morning which I thought was pretty harmless (and worked for me at that moment, honest!)

Anyway, I'm not in front of a machine where I can work on it, this second. But I'll roll back to the previous the moment I am. Probably in an hour or two.

I fixed a subtlish bug in Mind Traffic Control today.

I was trying to pull the email address out of user objects using user.email() ... however it seemed that you might be able to have users who don't have an email address and for who user.email() fails or returns None. Not common, I guess because user id's are based on Google accounts (which is mainly Gmail accounts I'd assume.)

Anyway I've now wrapped all attempts to get user.email() in a try. But next question, what should I do instead? When I find a user without an email, what should I use to identify him or her?

Ohloh is a YASN for free-software writers?

Stowe Boyd vs. the Neo-Fordians

Tim Burks talks about his language Nu (seems to be kind of Ruby-like behaviour with a Lisplike appearance (lack of syntax?)

Good references to Brad Cox's Planning the Software Industrial Revolution.

Cool. Didn't know this stuff before.

Dan Bricklin :

Socialtext is announcing today that they are adding integrated spreadsheet capability to their enterprise-level wiki, making use of the new SocialCalc code I've been developing with them. This isn't just a repository of separate spreadsheets, nor a separate standalone system like wikiCalc, but rather a full wiki where a page can be either the traditional paragraphs of text or a spreadsheet grid.

Cool ... now, if they could just *also* add network diagramming as another page-type that would really be getting somewhere. (Not, of course, that the grid pages in SdiDesk really achieved "spreadsheet" status ... but that was always a long term hope.)

Ah ... well ...

Actually, there may be some SdiDesk news soon ... you never know.

Update : Bricklin has longer background piece.

Reactored at GitHub

Great talk by Ian Piumarta.

From the fascinating looking Workshop on Self-sustaining Systems (S3) 2008.

Mind Traffic Control is ruthelessly useful. Indeed, it kind of strips out everything else *execept* usefulness.

Which is why, if you're thinking of trying it out, here's a hint. Don't try by putting in some random fake items like "test 1", "test 2" etc.

All they do is sit there and remind you to do them. You can't fiddle around organizing and prioritizing them, dragging and dropping or attaching little icons to them etc. (Which is the point.) So MTC looks inert.

In fact it *is* inert, because you don't really need it. MTC does nothing, because you want nothing from it. You don't really have a task of doing "test 1" or "try out mind traffic control" so the computer can't help.

It's as if Mind Traffic Control magically knows when it's being useful and when you don't really care.

So, to get the real MTC experience, put some real tasks in. Just 3 or 4 of the things you actually have to do over the next couple of days, and leave it open in a browser-tab on your machine.

Then MTC feels wanted. And the magic kicks in. ;-)

Aza Raskin dreams of easier Firefox extensions.

Let's hope that he's working on it.

Also, interesting to compare Steve Yegge on the Firefox vs. Emacs war.

Loads of nice stuff going on in Squeakland.

Showing Seaside demos to a colleague this week made me wonder whether I need to dive back in. It's been a long time (15 years) since I wrote any Smalltalk. And what with Python, Erlang, GeekWeaver + some C, PHP, Javascript; and Cache ObjectScript (don't ask!) in day-job, do I have room for Squeak?

But PyPy Goes Squeak looks fascinating.

As does this conference.

Folknology's Reactored is announced.

Nice ...

Goosh (Google Shell) ... not from Google, but damned cool.

Very interesting history of Nouncer.

GAE pricing looks reasonable from my perspective.

Or have I not thought through the scenarios? Glad I'm not trying to run some kind of video store of course.

Twoorl is free Twitter-clone written in Erlang.

The biggest problem I've noticed people having with Mind Traffic Control is with delegating. They were sometimes delegating without filling in the task description (thinking that the original task description would get copied across - which it wasnt't) or delegating to names of people which clearly weren't email addresses (or Goggle logins).

Now I've trapped this. When delegating a task, it's description gets copied (although you can still edit it), and the system also traps non-email addresses as targets to delegate to.

Keep watching ...

Om Malik suggests a freemium business model for Twitter that starts charging for tweets per follower.

Rather great Paul Graham essay on adult treatment of children.

This resonates :

Innocence is also open-mindedness. We want kids to be innocent so they can continue to learn. Paradoxical as it sounds, there are some kinds of knowledge that get in the way of other kinds of knowledge. If you're going to learn that the world is a brutal place full of people trying to take advantage of one another, you're better off learning it last. Otherwise you won't bother learning much more.

Very smart adults often seem unusually innocent, and I don't think this is a coincidence. I think they've deliberately avoided learning about certain things. Certainly I do. I used to think I wanted to know everything. Now I know I don't.

Update : and another good one on the internet as dangerous distraction for procrastinators. Twitter is ridiculously addictive. In the wrong state of mind I can find myself trying to refresh once a minute.

Mind Traffic Control is partly meant to try to harness that dynamic for good ... or at least for work. It gives you something to click manically hoping for new surprises to pop out, but it's only ever work which pops (assuming you only put real work or necessary tasks in).

Note, at some point I may import RSS feeds into MTC ... but remember, it's vitally important you do not connect MTC up to some kind of "news" feed. Don't drink from the fire-hose, as they say.

I realize there's no defined way of submitting bug-reports or suggestions for Mind Traffic Control. I need to set one up. In the meantime, use the comments for this post.

Couple of interesting posts on Twitter scaling issues.

Twitter starting a conversation with the community, and some speculation about the issues. It's not the writing it's the "view my page which collates from multiple other users" which is the issue.

If that's the case I wonder what an Erlangish solution would be. There's no real reason for the Twitter *database* to be centralized at all. (In this sense, Winer's thoughts on total decentralization are apt.) Could you not have multiple nodes, each with it's own database of some users. These nodes are going to have to talk to each other only when someone pulls a page out.

You could also, presumably, in the extreme case have entirely separate servers for different users, and only integrate the overview of all people being followed "at the glass" in the browser or RIA client.

Adobe Thermo looks interesting.

At first, of course, it looks just like another Visual Basic which is of note only because it targets Flex (and hence the Flash Virtual Machine).

But it's clever in the way that it imports Photoshop files and takes advantage of some of the logical structure.

Not for me, though, obviously.

Erlang behind Facebook Chat

A very natural conjoining ... spreadsheets become wikis.

Python as a Platform

Smart Disorganized Philosophy #2

Chatting with BillSeitz I found myself saying this :

MTC needs to have what Udel called the "special charmpower" of email ... that groups are spontaneous and form bottom-up from the flow, rather than having to be designed up-front, prior to the flow .... otherwise people keep using email

Another list view for Mind Traffic Control

OK, somewhat ugly, but they're there by popular demand ... full lists of your tasks.

Don't worry Mind Traffic Control isn't forgetting them. :-)

If you follow @exmosis on twitter you'll see he's busily trying to deconstruct Mind Traffic Control turning it into an untask-list or a barter-market.

@adrianh is constructively sceptical.

Lots here

Mind Traffic Control now has #tags.

If you use any word beginning with a # when describing a task, the task gets tagged with this word. And once you start tagging, a filter appears to let you filter your queue to present only those tasks which contain the tag.

Not entirely sure I've got the UI right on this one. It is simple, I think. But maybe not obvious, and maybe not in the right place at the moment.

Comments, bug reports, suggestions welcome as always.

BTW : a good thing about the GAE, it is so simple to add new functionality. This was about three hours of development and another 3 of debugging.

The reason for the longish debugging is what looks like a quirk in GAE.

At one point I was using a user object instead of a user email in a query. (I'm using user email's as user id in the database.) You might assume that this would simply fail with some kind of type error. But interestingly it worked on my local machine (somehow in the SDK it looks as though the user object gets correctly coerced into its email) but not on the appspot server where the query always returned nothing.

Read into that what you will.

Smart Disorganized Philosophy #1

I've been on a Smart Disorganized Individuals tip for several years, now. Always writing software compatible with that philosophy.

But what is the philosophy? What does this software mean?

In this series I'll start to make some specific notes towards that. Here's the first.

SdiDesk is wiki : a network of texts.

GeekWeaver is a programming language written in an outliner : a hierarchy.

Mind Traffic Control is a multi-user, dynamic queue, a "flow" of tasks.

Each is SDI. Each is completely different. Each is for specific purpose.

SdiDesk excels at capturing ideas and the relationships between them that are static.

Outlines excel at authoring or creating structure which is ultimately intended for a reader.

Flows excel at capturing change and movement.

OTOH, each is bad for something. These are true, even if you might imagine them not to be.

Wiki is surprisingly bad for authoring. Outliners are surprisingly bad for managing todo-lists. Flows are a surprisingly bad place to put ideas that you want to keep forever.

Wiki is great for writing, but awkward for the kind of reworking and structuring needed to polish a document for an external audience.

Outliners fail to match the dynamism of shifting tasks and priorities in the real world. They focus on making a structure of something which needs little structure.

MTC will lose your ideas when they are no longer in the future.

Today's Mind Traffic Control update.

The "defer" screen was simple, but not simple enough. Why make the user go through two mouse clicks - select a radio button and then the submit button - when this can be reduced to one? Now choosing how long you want to delay is a single button click.

I was also finding, when defering, that I wanted to remind myself of today's date. Now I've just added a display of today's date for convenience.

In general, I'm going to follow a philosophy of trying to identify and implement small improvements like this to MTC.

All suggestions welcome.

Just remembered Paul Graham's essay from earlier in the year. It really spoke to me (what with my bias towards the agile, experimental, piecemeal, wiki-natured, wabi-sabi way of design).

Of course it's my road-map for the ongoing development of Mind Traffic Control.

Day 2 of Mind Traffic Control in public.

Development continues with MTC exporting OPML

Mind Traffic Control roll-out continues.

Nothing spectacular in terms of numbers, but friends are making interesting comments and suggestions; and some strangers are discovering it via the Twitter announcement.

Eikeon pointed to twemes as a place to search for the #tag #MTC on Twitter.

Good chat with Folknology. He's encouraging me to open up the data. Hence I did CSV export of your tasks. Al, likes RSS/Atom but I still need to find out whether feed-readers can do the required Google login to get at a non-public feed.

Rup3rt weighed in with a number of comments starting :

Wow an invisible list machine almost like a gumball/toy dispenser ...

and continuing

I like to see what is ahead but it is good with the delegation - I can line up things for the estagiarios. This is great for non-negotiables like driving instructions, recipes.

* The font for "This task was delegated to you .. may report the sender for blacklisting and delete this task" is a bit big

* I would like a difficult autoaddress (1gg23ff3d3reedd@mindtrafficcontrol.appspot.com) to delegate tasks by email and maybe I could get other tasklists to feed into MTC using this (to just top it up with priority1 or todotoday tasks)

* Maybe paste a list of (line break delimited) tasks that are automatically added

* Maybe show in a widget

* Maybe show total number of tasks

* How would time estimates of task duration fit in? It could be a diversion to see "You have 13 hours of traffic to negotiate")

Good suggestions. I'll be interested to know what everyone else thinks.

But here are some initial thoughts :

Font issue, yep. Mailing tasks in to Mind Traffic Control would be great; a very useful addition. But I'm not yet sure I can do it on the Google Application Engine.

I might be able to do it using another hosting service but that's introducing a lot of new complexity into the system : running two different hosting services, writing scripts on the old-skool one to pass emails it receives into MTC in some way, being responsible for security (how do I stop spammers mailing tasks to you via the system?) etc. Going to take a while to figure that one out.

Importing other formats and / or entering multi-line, multi-task lists is easier and more likely to appear soon. I'll definitely go for the multi-line input box in the short-term future. I'd be interested to hear what formats readers currently use for to-do lists so that I can import them. I immediately think OPML, of course. But are there others?

Widgets ... actually I've been playing with OpenSocial and Orkut Widgets for my other GAE project. It's cute ... but I'm not yet sure what I'd put in a MTC widget.

Task counts and time estimates I'm ambivalent on. I don't want to make MTC a paranoia inducing space where you arrive on the front page to be presented with something saying "Welcome to Mind Traffic Control. You have 280 outstanding tasks". On the other hand, I see that people would like some sort of overall measure of what they have to do. More thinking on this.

Finally, there's a couple of issues of "philosophy", particularly why I'm resistant to letting you flag tasks with "priorities" but leaving that for the next post when I'll be more coherent.

Softlaunch ... no fanfare ...

Mind Traffic Control

This is my first Google Application Engine application. A kind of next-action list for the Twitter generation.

Basic instructions :

1) You need a Google account to login (eg. Gmail) ... it's a GAE app.

2) There's a blue box for inputting tasks you need to do. Once you start entering them, Mind Traffic Control will show you your next action.

3) At which point you have a choice. Say you did it. Delay it (push it to the back of the queue). Delegate it to another user. Defer it to a future date. Or delete it altogether.

4) Keep going.

There are a couple of ideas behind this. First, that people have been getting too hooked up on the idea of "organizing" their tasks and actions into lists. But organization itself isn't really an end, it's a means towards making sure you remember and discover your tasks, as a means towards the ultimate end of getting stuff done. So instead of thinking of tasks as list-management, Mind Traffic Control treats tasks as un-structured flows.

The real focus is on capturing the tasks as quickly and easily as possible. So, inspired by Twitter and Folknology's Rel3, the input box is always there at the top of your page.

The other focus is when the task is presented to you. At this point you need all your options to quickly route the task. In a more static system we might say route into particular "bins" but here it might be better to say "channels" (unless it's the ultimate sink of "done" or "deleted".

Another idea ... it's meant to be the simplest workflow that could possibly work. Just short messages which can be routed. Everything else has to be constructed by humans on top of this. But as, once again, we've seen with twitter, it looks like humans are pretty good at filling out these communication channels.

Anyway, off to bed ... will write more posts here soon ...

please have a try with Mind Traffic Control, and if you have any comments, feel free to post them here.

I've now been playing with GAE for a couple of weeks, and it's still looking pretty good. I've now got an Orkut developer account so I can make Orkut widgets tethered to a GAE based service.

Meanwhile looks like GvR has been doing some cool stuff within Google with internal development tools, and has just released it to the outside world, thanks to GAE giving us access to BigTable etc.

The change in mindset needed for GAE

Zed Shaw's take on GAE is entertaining.

Now this is very cute :

[(f.title,f.key()) 
    for f in db.GqlQuery("select * from Film order by title")]

See that? A GQL (query language for GAE) query object automatically supports iteration, so you can use it in a list comprehension.

OK ... better start finding out about Google App. Engine

Damn! Don't think I was in the first 10,000 applicants for a beta ... got the downloadable development environment though ... watch this space ...

Update : Actually, I did get one ... :-)

Ruby on Javascript VM?

Sometimes you need to take a step back from a problem, to see the forest for the trees.

GeekWeaver was being held up by something that seemed a rather complicated knot; one that I've wrestled with a number of times but never really untangled to my satisfaction.

So last night I decided to restart with a clean slate. I didn't look at the existing code or existing unit-tests. Didn't even open my IDE or project file.

I just started up IDLE and in a single file, redid the whole thing from scratch, in a fast test-driven stylee.

Amazingly I think I've come out with a cleaner solution, messing around with fewer classes and no attempt to over-re-use with inheritance. The whole thing took about two hours total. Of course, it's not quite finished, and still needs to be integrated back into the main codebase. But I like it a lot more. It's shorter and easier to understand.

Sometimes (at a certain granularity) old code *is* more of a burden than an asset. Understanding what it does and what you can do it is a cognitive cost that outweighs its value. Don't be afraid to throw code away when you need to.

Caveat for small granularities. Obviously rewriting an entire application from scratch is a different matter.

Sweet Expressions ... great idea for making S-Expressions more like other, more readable programming languages

without losing their power (such as generality, macros, quasiquoting, homoiconicity, and easily-manipulated program fragments).

Hat-tip Folknology

Wow! I need to get into Buildout.

Update : Good blog-post accompaniment

Very interesting. Recently my "ProgrammingInWiki" page on WardsWiki seems to have come back to life. Along with a companion WikiIDE.

A WikiDE is, of course, something I've been working on in the background for some time. But I'm far from having anything to show yet. So if these guys get there first, good luck to them.

I just want to be able to start working this way.

Damien Katz has more interesting comments about Erlang's "ugliness".

(hat-tip @Folknology)

I've been wondering what Ward Cunningham's been up to at Eclipse. Jon Udell is on the case : Interesting tools for software developers.

Apparently's Ward's now at a wiki-startup. (Mark Dilley seems to be involved too.)

Cobra seems to be like Python with optional compile-time type-checking and other quirks removed.

It also has contracts, which ensure certain conditions when calling a function. However, wearing my new Erlangish hat, I wonder what contracts buy you over pattern-matching arguments?

Unit testing built in to functions is interesting. Not sure what I think of that .... yes, it's useful and convenient. But a) it may clutter up clean code (I tend to let a certain amount of messiness leak into my unit-tests, that I don't let into my code); b) it's not so easy to see how it can be used for test-driven development - how can I write my test before my class? And, worst of all, c) I tend to use unit-tests when refactoring and moving functionality from one class to another.

It's very simple to change a :


x = MyClass()
self.assertEquals(x.f(4),10)


into


x = MyClass2()
self.assertEquals(x.f(4),10)


and have my unit-tests already defining the new target I'm aiming for.

But if the test is is actually *part* of MyClass, then that's going to require more premeditated moving stuff around during my experimental refactoring.

I'm way too busy now ... but I'm having some kicking ideas about my own grandiose bid to improve programming ...

... think GeekWeaver in an SdiDesk-alike editor (obviously).

Then imagine that *everything* is a template.

That's the way that GeekWeaver is already going. Everything is a template (ie. has named slots that can be filled). Function calls are just the injection of a data-block into that template. Another way of putting it, all objects know how to handle the "call" message with a data-block as argument, even if they don't do anything very useful with it.

But now imagine that all the types of things you can get in SdiDesk ... text pages, grids, network diagrams, are also templates. And you can plug and pipe them together any way you like. One page can hold a table, another a network diagram-shaped template, and a third can be specified as the result of injecting the first into the second.

Hmmm ... this definitely looks like it's going in the right direction.

Somehow I feel I should be exploring WikidBase

Looks clever ... and in Python.

Here's a question ... why is Erlang so ugly?

I don't mean that in a pejorative way (not much, anyway). I mean, I really love what it does. I'm totally impressed with Erlang's power and simplicity. I'm writing simulations which are about a quarter of the size of the Python equivalent. So this is not to be taken as a criticism of Erlang which I'm definitely committing myself to, this year. Rather this is some random speculation about programming language aesthetics.

Erlang is wonderfully concise. And yet, somehow, unlike Python, unlike Haskell, it just doesn't come across as beautiful. It's confusing. It looks cluttered.

A couple of lines look fabulous. But the simplicity doesn't scale.

At first guess, there seem to be three issues.

a) As people have noted, the record type is ugly. It is. And counter-intuitive to use in patterns (although I may just be stupid).

b) In general I think it's good and brave thing to take a stance *against* objects. But I haven't figured out how to do encapsulation (data-hiding, abstract types)

Sure, I can define polymorphic functions (one clause at a time per input type) which is a lot shorter and more powerful than overloaded methods in Java. But it has the effect of jumbling all my data-types together. Which just feels *wrong* to me. (Of course, maybe that's some residual OOness in my thinking.)

But I think that may be part of the bigger issue :

c) erlang doesn't seem to have resources for "programming in the large". And, ironically, because erlang is so powerful, that problem becomes visible at a smaller scale - precisely because in erlang "large" programs are actually "small".

Or rather, the only resource is "modules" (which means breaking up into multiple files - always an extra overhead.)

But if you avoid breaking things up into files, the opposite problem becomes apparent. I can do the equivalent of a small Java class (let's say something around 50-80 lines) in about 6 to 10 lines of Erlang. But 10 lines of erlang is too small for a separate file. So I'll add the next 10 line packet of functionality to the same file. By the time I'm up to 4 or 5 packets that would be handled as different classes in Java or Python I may have written only about 50 lines of code ... but it's all running together!

There's no higher level of organization to distinguish and separate the code. In Python I often put 5 or 6 small-medium classes in a single file or around 300-500 lines. But the indentation and editor make these reasonably distinct and identifiable. In contrast, my equivalent 200 lines of Erlang have no visual cues to break them up. I can't use functions as a visual element because pretty much every line is a new function (except when I'm doing I/O, which has its own "issues").

I'm left with using comments but my editor (Komodo), excellent in many ways, doesn't actually know Erlang and so doesn't colour them differently. And, in general, because functions are powerful, they *are* short : one or two lines. But those lines are typically much denser than Python. Even if a dedicated editor would colourize them, I'm not convinced that's such a big win at this density. On the other hand, I don't want to artificially scatter them out into multiple lines. I'm not trying to recreate Python with a slightly less appropriate syntax. I want to take advantage of Erlang's power and conciseness.

But I wonder what the right aesthetics for a language as high level, dense and abstract as Erlang is. Haskell looks cleaner to me - maybe because it does abstract data-types right. But can it solve the problem of organizing your larger-scale things? Lisp is no role-model. ML and its offspring have always repelled me visually. (Nothing can be more ugly, dispiriting and patronizing in a programming language than an explicit "begin" statement.)

Suggestions, anyone?

And I thought I was ambitious, trying to write a programming language!

My friend Oli has decided to reinvent programming as we know it. Details are still trickling out via his web-site : Semantic Programming. And I'm in frenzied skype conversation with him, trying to figure out what it's all about.

In outline, it starts from some intuitions behind the Semantic Web : that there should be a massively parallel, distributed graph-shaped database of facts (relational assertions) represented on different machines across the world. But it then layers programming on top of that. Instead of a passive data-structure crawled by scutters, SemProg agents (roughly, the servers which manage different data-nodes) are active. There is message passing between the facts themselves, and agents may have hardwired interpretation to act on some facts (what Oli is calling "axiomatic" understanding), or a "deductive" understanding (I guess rather like Prolog inference), and even a "behavioural" understanding via (I guess again) learning from observing other agents.

I'll keep following this here on Smart Disorganized. Very interesting if it works out.

Would GeekWeaver support Semantic Programming? It seems like Oli is thinking of multiple editors for different types of information, all of which compile down to the same underlying graph-structured format so that the data can be combined. (Rather like Language Oriented Programming.) It seems quite possible that GeekWeaver could output something like his graph-format. I'll certainly be experimenting.

I'm also trying to persuade Oli to look at Erlang as a potential implementation language for the distributed virtual machine. I'm increasingly impressed by Erlang. Finding it very powerful and concise.

Towards End-User Programming with Wikis

Paul Graham's Arc Challenge . Examples in Lisp, Smalltalk Seaside, Ruby, Perl and Python (using generators instead of continuations)

And an Erlang response.

@adrianh suggests looking more at Parrot Compiler Toolkit ... which is very cool.

I'm finding that language design is hard. Not just the implementation part (although that's hard too) ... but just figuring out how to make a syntax that allows all the things you'd expect to be done in an elegant way.

I'm starting to appreciate how clever certain common design patterns are in language design, and how hard it is to deviate from them. More on this soon ...

Ian Bicking on what PHP deployment gets right.

I'm playing with pipelining in Python over on Composing.

Here's an interesting question (after reading this).

Why does anyone want to write "large" programs? Everyone knows large programs are difficult to build and maintain. Want we want is *small* programs that happen to be powerful (do a lot) and scale over large numbers of users. Is there any reason to think that "large" program correlates with either of these things?

Supplimentary question. Are 10 programs of 1000 lines each, which have to nevertheless interact (via pipes or XML-RPC calls etc.) actually any easier than one 10000 line program?

Update : Folknology has been making a pretty decent case for Erlang as a language optimized for writing software as a swarm of small programs.

Discipulus, patterns for programmers.

LOLCODE IZ IN UR COMPUTR RUNNG THINGZ

Python on Silverlight?

I'm intrigued despite myself.

E.E.Kim on tools

Looking at some slides about Perl 6(pdf) ... some cute features.

Quick note : Blahsploitation is no more ... long live "Composing"

The impotance of being earnest :-)

Update : which reminds me of a discussion I'm having with Sig

Update 2 : Seems like a lot of ongoing discussion around the place. Joel Spolsky explains.

Update : for people wondering how the whole new-SdiDesk-in-Adobe-Flex? thing is going. I solved something I thought was a problem yesterday.

I now have a (very fragile) Flex front end which can pass plain-text GeekWeaver programs to a web-server with GeekWeaver embedded, and get a compiled chunk of GeekWeaver out.

That's very cool ... unfortunately it also revealed yet more problems with the GeekWeaver compiler which need fixing before I can release the next version.

Still, it looks promising. The signs are better and better that something interesting is coming out in the next 3 months. :-)

Restless Mind Syndrome (propaganda for the ultra-cool looking LiveScribe)

He he! Perhaps NooRanch is NooVictorian ....

(I know, bad, bad pun)

What's this?

• 1
• 2
  • 21
    • 211
  • 22
• 3
• 4
  • 41
  • 42
  • 43
    • 431
      • 4311
• 5

An HTML list that came out of GeekWeaver, when I called this recursive function :

::rec
  .<ul>
    :for x,, #__
      .<li>
        ${x/=}
        :hasChild x
          :rec ++ #x__

on a chunk of OPML.

That's pretty much how it's going to look, folks.

::rec defines the recursive block (or function)

.<ul> creates the UL tag

:for x,, #__

is creating a loop through all the anonymous (unlabelled) children of the tree which is being passed to the call. (This list is represented by the symbol __)

Note that I'm trying to keep GeekWeaver as functional as possible so there's no real "assignment". x is scoped so that it only exists inside the block of the for-loop, it's bound once before the block is evaluated, but can't be updated after that.

${x/=}

This is the new way of accessing variables. x is actually bound to a sub-tree (Almost everything in GeekWeaver is a sub-tree, except a couple of weird cases like __ which is a list.)

Although a simple $x still works if you want the whole of a tree flattened into a string, ${} expressions give you a way to pull data from a single path in the tree. In this case we're just getting the text content of the root node of x.

But it's also possible to write stuff like this : ${company/employees/__3/name} which means from the symbol "company" get the labelled child "employees" from which get the non-labelled 3rd child, from which get the child labelled "name".

:hasChild, like :for, is a build-in function. If the tree in x has any children, it evaluates the body of the :hasChild tree. If not, it returns nothing. This tests for our "base-case" when we've hit a leaf of the tree we're recursing through.

The final line :

:rec ++ #x__

is the one which I've been struggling with for the last three months, ever since I started trying to figure out how to support recursion. I'm still not 100% happy with the solution I've come up with, but it's getting there.

:rec is calling the recursive block again. And obviously I need to pass the children of x as arguments. However the way the function is written, it is working on the anonymous children (inside the default variable __)

How am I going to get the children of x into the __ variable inside the next call of :rec? That's what ++ does. Not sure what I'm going to call this, I may call it a "pivot" although that may confuse as much as help. It captures something of the idea that I need to swing the anonymous children of x (#x__) around so that they can go into the next call of :rec as though they were children of the calling node

Like I say, don't know if I like this name or the ++ symbol being used for it. So consider both as provisional for now.

OK, I'm too tired to make the build with this stuff working tonight. I'll try to get a build together in the next couple of days, but meanwhile, if this looks interesting to you and you can't contain yourself, send me an email (interstarATgmail.com), say hello and I'll see what I can do.

Wasabi part 3

And some of Martin Fowler's book on DSLs is online.

Hey! I like NeoVictorianism.

There's a movement I can sign-up for :

Built for people

Built by people

Crafted in workshops

Embracing irregularity

Inspired

Just spent the last few hours downloading and playing with the beta of Flex 3, Adobe's IDE for Rich Internet Applications (ie. applications running on the Flash Virtual Machine) which is based on Eclipse and has an XML-based UI / form description language more or less like HTML.

I'm having two thoughts about it. One is a kind of sigh of relief. This is, after all, finally, The One. After years of fruitless searching I'm pretty sure here's a framework I can settle down with and commit to, and start making babies with. At least, it's more or less mature enough, handsome enough and well endowed enough to put these thoughts into a girl's head. A browser and a desktop? Holidays in Windows, Mac and Linux? Own grid and canvas. Tabbed notebook and some cute chunky buttons.

And it's all done in a way that's pretty self-evident when you look at a few examples. Forms and input widgets are described in XML. They layout nice; and you can start banging them in and prototyping the look of your interface in a couple of minutes. The round-trip from coding to running and testing is a bit slow on my poor 512MB Vista laptop, but it's going to be bearable. And the Eclipsyness of it all is comfortingly familiar if a trifle overblown.

As long as my next experiments turn out right (the one where I try to find a tutorial example of pulling data off a server over http, and the one where I try to compile with AIR into a stand-alone application) then I'm sold.

But there's another part of me going, "huh? Is this all there is? WTF?"

I mean, it's 2007 and I'm happy because I've finally found a way to make GUIs that's sufficiently lower than my pain threshold that I might actually get a piece of software released again. Zowie! But that's what I had with Visual Basic 6 - which came out in about 1997!

In fact, I was already a Pythonista before I started writing SdiDesk in VB. And I only pulled out VB (a language I thought I'd left behind for good) because I got impatient to see what the UI of an SdiDesk could look like and thought I'd prototype it. As often happens, the prototype spiraled out of control as I kept thinking, "maybe I can just also add ... " and within a month or so it had already started to grow into a real program. Another phase of development with some serious refactoring and cleaning up the internal architecture, and it was a quite respectable and powerful bit of software (If I say so myself; I'm talking about the time I made the tutorial screen-casts.)

(Then, of course, I hit the crisis of not wanting to be on the Microsoft treadmill and forced to upgrade to .NET; even though it was obvious that VB6 was as extinct as a very extinct thing from the Lower Devonian period. But also of not having any viable alternative. )

So there's a sense that Flex smells extraordinarily similar. I can see how you can knock out your prototype interface and start building backwards from it. That feels good. That's why it seems like this is plausible option to get development rolling again.

But, like I say, it *is* basically what I had 10 years ago. Except with javascript dressed up in Java's suit and tie to look more grown-up and respectable. And XML.

In fact, I was gchatting to Zbigniew earlier today, and realizd that all of this stuff is hardly a big advance on Hypercard back in 1985. Or perhaps Smalltalk 1972. Why the hell haven't we progressed further? Why am I struggling on each new platform to rediscover the level of comfort I had on the previous one? What's going wrong here?

I suppose it could just be that the idea of quick GUI builders is inherent in the idea of a GUI?

Or maybe we programmers of the noughties need to get our acts together and start coming up with serious new, cool shit. Stuff which couldn't have been thought up in the 60s and 70s. Stuff which is radically easier and more productive than something we had 10 years ago. Something like reinventing Lisp with a cleaner (non)syntax ... erm ... well anyway, I'm off to do more experiments with Flex and try to get it to talk to some kind of server.

If I succeed, then expect to see some interesting developments along the lines I mentioned earlier today ... steps towards a new SdiDesk, possibly a GeekWeaver development environment ... maybe even the long fabled, but never released SystemSketch. Or even the more outré things I've got buzzing around in my fevered imagination like "SexyCells" and "FlowerBrush".

Of course, it still sucks that Flex / Flash seems to have no musical ability whatsoever so Gbloink! doesn't look like an option. Which is a double pity because I think it would make a great Chumby widget and that would have justified me buying one.

:-(

37 Signals on why enterprise software sucks

Quick Note : I just had a revolution in my thinking, triggered by Enso but influenced by several other recent trends.

You write Enso "extensions" as XML-RPC servers sitting on your local machine, register them with Enso and it calls them using XML-RPC. I tried the example from the tutorial and it's very cute and simple. So I've decided this may be a way forward for the dilemma which has kept the "new SdiDesk" on hold for several years(!!!)

If you remember, the issue has been whether the new SdiDesk should be a "web"-application (accessed through the browser) or a desktop application? And how to implement the UI.

The problem with the "browser" answer has always been : "but how to do the network diagramming bit?" - which requires interactive vector graphics. SVG doesn't seem quite stable or cross-browser enough. Canvas isn't cross-browser. Flash is proprietory.

The problem with the "desktop application" is, well, the many rival Python GUI libs with different degrees of maturity. And the question of getting them to work cross-platform.

A meta-problem : I just don't have time and energy to go through learning lots of different GUI frameworks to decide which I want to use. And the ideal answer to the browser / desktop question is probably "both" which doubles the amount of work.

In fact, one of the motivations for GeekWeaver is to see if I can use it define a higher-level UI description that can be compiled down into both XHTML and a GUI widget-set.

Of course, with XUL, Open Laszlo, Adobe Flex, XUML etc. lots of people have been looking at something like HTML for defining desktop apps. too, but the free Python frameworks don't seem to have caught up with that yet. XUL sounds most promising especially with the open sourcing of Active State's Komodo ... but that's also a big complicated thing to even start looking into.

Anyway, inspired by the Enso extensions and Bruce Eckel's interesting example of hooking up a Python XML-RPC server behind an Adobe Flex application I've been wondering about blowing the project apart entirely into a number of services connectd only XML-RPC or something more ReSTful.)

So there'll be a WADS (SdiDesk PageStore) service, a GeekWeaver interpreter service, a "user navigation history service" etc. Even the glue which ties all this together will be another service that's neutral about the user-interface. Effectively the model, view and controller will be completely separate programs. (I know, I know ;-) )

Then there'll be a variety of different types of access with different UIs.

Now that Adobe's AIR has put Flash on the desktop and made it a serious rival to the Java virtual machine - I *am* tempted to put some time into learning it. It would give me a reasonable GUI widget set (including TreeGrid, yay!) and the vector graphics needed for networks. It will run both on the desktop and in the browser and on Linux, PC and Mac. The tools are free-as-in-beer (Flex beta) or free-as-in-speech (Open Laszlo). It's also possible to imagine generating the Flex XML or the Open Laszlo format directly from GeekWeaver. (Aside : I wish ActionScript hadn't borrowed quite so much from Java, but still, I'd only be using it for the front-of-the-front-end UI stuff.)

Then I want to experiment with Enso access - for example, Enso commands like "page HelloWorld" or "pagehistory ChocolateCake"

And there'll be web-browser access probably through a "gateway" that accepts http requests and spits out HTML for a browser. The only thing I don't know is whether I *really* should be using WSGI somewhere here.

Anyway ... that's a quick update of my thinking this week ... tune in soon for the next GeekWeaver release (really, it's coming together, and gonna be very powerful). Then it's back to work on this larger project.

Next installment of the must-read series on FogCreek's Wasabi.

Fascinating to see the problems they've come up against and the solutions. What Stefan calls "picture functions" sound close to GeekWeaver "blocks".

Cool RedMonk TV on open source startup.

Over at Platform Wars I'm tracking Google's "square circles" (hierarchical tags)

Very interesting ... seems that FogCreek's Wasabi is a language to compile into the various parts of a web-app in ASP or PHP.

So kind of a competitor to GeekWeaver. :-) Better take notes.

I wonder if it's gonna be released to the public.

DAn McWeeney has a great post on what he calls "synthesizers" : which seem like generalist / hacker / blogger / communicator type.

Personally I think the existence of the web is pushing us all in the direction of being synthesizers.

This story about rejecting Rails in favour of PHP is very interesting to me now I'm on a PHP-tip.

Some good links on Diseconomies of Scale ... via comments at Bruce Eckel's

Important statement (actually essay / braindump) on my thoughts on widgets and YASNS over on my personal blog.

Mark Bernstein has a great post (and sounds like a great talk) on NeoVictorian Computing.

Wants to return software development to the personal and to making meaning.

Via Bill Seitz

Cool!

Thanks to Dave Stein I now have "programming" on my Behance Profile

New meme : GeekWeaver is Language Oriented Programming for everyone else.

Good weekend for GeekWeaver development ... you can now pass arguments to functions that are more or less table-shaped - like this :

:f Fruits
  apples,, oranges,, pears
  grapefruit,, passion-fruit,, grapes
  bananas,, lychees,, mangos  

It's not public yet, but it will be available in the next installer.

Also, although GeekWeaver is designed to be written in an outliner, I'm close to a plain-text mode which will look something like the above.

But I have a problem which I'm still not sure of how to solve. GW has to preserve literal text. So using spaces for indentation like Python is not so easy.

abc
  xyz

could mean either a node "abc" with a child "xyz" or two nodes at the same depth : "abc" and " xyz". How can we tell the two apart?

Currently my plain text parser works with code that looks like this :

* abc
** xyz

Which keeps the continuity with wiki-markup and disambiguates from

* abc
*   xyz

But it's pretty ugly if you really were going to do any serious programming with it.

Suggestions welcome.

Thanks to Zby for turning me on to Behance which seems to be an interesting combination of social networking for creatives backed by a GTD-style how-to-organize methodology for disorganized creatives, and has special stationery too. Yay!

Could fall between a lot of stools, or could be the next-big-thing (more fun than Linkedin, more serious than Bebo, easier to use than degenerate art)

I do like the explicit emphasis on trying to solve the problems of disorganized people.

I don't like that "software architect" is a creative label you can attach to yourself but "programmer" isn't.

Amusing : job vacancy advertised in Twitter.

That's how evolved the coral reef is getting.

Just commented on a topic dear to the hearts of all smart, disorganized individuals over at Ward Cunningham's Vision Thing.

Resolver Systems seem to be a cool, python + spreadsheet company based in London - with the right attitude. Nice demo.

Hmmm ... and they're hiring ... (drifts off into daydreaming)

Joel Spolsky :

And your programmers are like, jeez louise, GMail is huge, we can’t port GMail to this stupid NewSDK. We’d have to change every line of code. Heck it’d be a complete rewrite; the whole programming model is upside down and recursive and the portable programming language has more parentheses than even Google can buy. The last line of almost every function consists of a string of 3,296 right parentheses. You have to buy a special editor to count them.

Spooky.

I mean, that wasn't exactly the whole GeekWeaver gameplan. But "high level" lisp-like language that compiles down into complex web-applications, is not so far off. I was thinking of Dojo as the Javascript library, PHP at the server, and Facebook itself (or Ning) as the layer at which applications can be glued together. So read with Marc Andreesen too.

It's the zeitgeist I tell ya'

I'd forgotten that Processing Processing by Paul Ford was one of the inspirations for GeekWeaver.

Reading more discussions on the business of Joel has got me thinking about the problem of marketing GeekWeaver.

So here's the pitch ... in the classic 5 points of selling ;-)

1) The pain you didn't know you had!



Oh, the horror, the horror, too many tags too many files, you have to reach for the mouse and click to find the file menu and open up a new file and then scroll around in that silly little common dialog where the files are all represented by little yellow blobs that are 2mm square and then you click one and open and your hand has to come off the mouse and down to the keyboard and you try to scroll around but there are tags and tags and tags then you change something and you forget to close the tag and everything's position gets screwed up and you think ... oh if only I'd used CSS, but there wasn't time and anyway you have twenty pages the graphic designer made by copying and pasting in Dreamweaver (oh BAD Dreamweaver) and now the client wants them all changed and if you were working with a content management system you'd have put some common things into includable subtemplates but you don't have that and you think why am I watching the grains of sand draining through the hourglass of my life doing the same mechanical transformations on dozens of static html and xml files and you know the customer will only want them changed again next week .... aaaaiiiighhh why can't html be like a proper programming language instead of this ???? WHY GOD? WHY??



2) The generic solution.


Yes, a programming language. That would have the requisite controls over abstraction that I need. And a decent IDE. If only writing HTML files could be more like a programming lanaguage.



3) Our solution is your solution.


Fear not because GeekWeaver will solve all your problems.

All your pages are kept in one file, which you can navigate around with the keyboard in the comfort of your favourite OPML editor. You never have to close an HTML or XML tag again with the .<tag> notation that takes advantage of the outline structure to know what blocks are meant to be inside what blocks. Never, ever copy and paste when you can define your own re-usable, parameterizable blocks at any level you like including outside the level of a single page ... that's right folks, you can define a reusable component which actually evaluates to a number of pages or other output files.



4) Accept the price


Think ... just think ... of the hours of your time that could be saved if your static web-pages (and everyone has a few tucked away somewhere - remember that old departmental intranet-site no-one's been updating in years? Remember that customer who needed to access the documentation off-line?) were all safely in GeekWeaver format? How much would that be worth? (And at 50 quid an hour you're underselling yourself, but hours of tranquility can be bought for a mere ... well, let's see download and install Python 10 mins on Windows, probably already have it on Linux and Mac. OPML editor? 5 mins on broadband. GeekWeaver ... 2 mins?



5) Act now!


Get it today! because it's Saturday and you know you'll be too busy wasting time manually editing HTML once Monday comes around. And if you're an early adopter, and send me emails, then I'm more likely to listen to your requests and stuff ...

An idea I had, over at Joel's Business of Software :

Here's a thought.

I was re-reading this a couple of days ago :

http://www.joelonsoftware.com/articles/fog0000000348.html

as I'm trying to persuade the overworked developers in the place I work to take some time out for this kind of cleaning.

And it occurred to me that there could be a market for specialist code-cleaners. After all, you have specialist office cleaners instead of making the clerical staff take the Dyson around after work. Why not somewhere you can outsource legacy code to simply be carefully and lovingly refactored and polished without any claim that this company are doing new development?

Normally companies from rich countries imagine they can bang out a few UML diagrams, ship them off where labour is cheaper and have it coded up. We all know that this is a fallacy (coding and designing are too tightly entwined).

But could "take my OK but ugly code, with these existing unit tests, and return it with identical behaviour but all the HTML tags closed and the variable names consistent" be a more viable way of dividing labour?

Anyone know of anyone currently offering this? Could "code cleaning" be a small service business? Maybe even part-time for people who have a few hours free each day to work from home, know how to code, but can't commit to a full development job?

I've done a slightly improved GeekWeaver screencast.

Still very sketchy, and without audio, but the SWF is smaller, and shouldn't crash.

It just concentrates on the basics :

* making several pages
* using the .tag notation with outline indenting to save on closing HTML / XML tags
* defining a reusable and parameterizable block (in this case, page template) and calling it for each page.

I'll try to get the next one done in a few days, with some more surprising capabilities.

I've been struggling with Python's distutils over the weekend, and I'm finally making some progress. There's a new GeekWeaver page and a new release.

The page is part of a revamp of my personal site, which is now being built with GeekWeaver. And the release is the first built with distutils.

I was hoping that it would put the code somewhere on the PATH so that you can include the GeekWeaver modules from anywhere, but that doesn't seem to be happening. So although the code goes into "site-packages" I'm not sure what that buys you at present. Any Python experts able to shed some light on this?

What's definitely happened is I tidied up the whole directory structure. Unit tests and library modules are in sub-directories, and you can run the program from wherever you like as long as you give the right path to the gwMain program.

If you go into the "examples" directory in this new distribution, you'll see some examples of .bat and .opml files. You should be able to run the .bat directly there. You can put your own projects in their own separate directories using this as a model.

As always, I'm keen to have feedback, questions and suggestions for GeekWeaver. It's starting to come together although I also need some proper docs and screencasts to really show what it's capable of. The example files show the most basic (but most useful) features of GeekWeaver, the fact you're using an outliner for your whole site, the fact you can manage multiple pages within the same file, the way you can use .< notation to save yourself having to close tags, and the definition of re-usable, parameterisable blocks.

There's a more complex example of GeekWeaver's power in the tests sub-folder. It's not that well explained yet, but it shows how you can produce tables, how you can map a block-call to a sub-list of items, define blocks which are bigger than a single page, and pass block-names as parameters to other blocks.

I see Zbigniew beat me to blogging about wiki-wednesday. That's one of the perils of laying around in bed all morning.

He's dead right about Alan Wood's presentation being very exciting. I'm probably even more impressed by Wood's "Rel3" than Zby is.

What most impressed me was Alan's apparent good taste that seems to be formed by real experience. Rel3 isn't an attempt to take a particular genre of social software as it exists and force it into the enterprise, but a crafted, pragmatic response to particular needs.

Yes, it looks very like a non-threaded discussion forum. In fact it's also a bit like a page of "River of News" for a single issue being handled within a group. There are echoes of TopicExchange and explicit inspiration from Twitter.

Behind the scenes things seem just as interesting. Written originally in Java, then ported to Rails and now to Erlang. It's designed to use Amazon's S3 and EC2. (The fact that it keeps data in files on S3 prompted leaving Rails.) Each page (or conversation) allows files to be attached so you can manage Word, Excel, PDF docs, etc. When asked if he's thinking of creating some kind of Ajaxy online editor for collaborative documents, Alan sensibly says he won't re-invent the wheel but will provide hooks to other providers of this kind of stuff.

In other words, this is a very nice, minimal "glue" to tie together the people and resources that temporarily need to be brought together to solve a particular problem. After which the page doesn't hang around cluttering up your ontology when you have lost interest in it. (I mean, it probably *does* hang around. But Alan Wood, interestingly, contrasted Rel3 with wikis which allow a small group to work producing a more permanent document that must stake its claim to some part of the name-space.)

In this, it has some of the best features of other social software, in particular, I think it has potential to rival email's special power of spontaneous group forming and dissolving. Let's see if actually creating, subscribing and unsubscribing to pages can be quick enough.

Also, Alan has a very smart blog at Folknology