Comments for Untyping

Comment on Libraries update by Noel

Noel — Wed, 10 Jun 2009 09:20:03 +0000

Thanks for the bug report, Lori. We’ll look into it.

Comment on Libraries update by Lori

Lori — Tue, 09 Jun 2009 21:42:45 +0000

Do you have plans to update the sqlite3 support in Snooze?

Snooze currently depends on (3 4), which unfortunatly bombs out on the latest PLT Scheme with the following error:

../../Library/PLT Scheme/planet/300/4.2/cache/jaymccarthy/sqlite.plt/3/8/sqlite.ss:12:3: expand: unbound identifier in module in: ptr-ref

The latest version of the Sqlite library (4 2) is fully compatible with the latest PTL Scheme… but unfortunately it can’t be used with Snooze due to handling its own type conversion.

Basically, the latest sqlite library will return an integer instead of a string, but snooze will try using string->integer on it anyways.

Comment on More State on the Web by Noel

Noel — Fri, 27 Mar 2009 11:34:30 +0000

There aren’t any significant issues beyond the usual web issues. You can encrypt continuations and store them on the server (on disk, say) if security is especially important. Forging a continuation is as likely as forging a session ID in other frameworks.

Comment on More State on the Web by Jay McCarthy

Jay McCarthy — Fri, 27 Mar 2009 01:20:08 +0000

I think it is worth the price, because you can always also convert the other module and in my conversions, I’ve never even needed to do that except with a handful of list combinators (like map). Obviously your mileage could vary, but it is certainly better than not having continuations at all and having to it all manually.

Comment on More State on the Web by dysfunctor

dysfunctor — Thu, 26 Mar 2009 14:00:29 +0000

That’s cool, but the requirement that continuations NOT be in the context of another module seems burdensome. Is it worth the price?

Comment on More State on the Web by Steve

Steve — Thu, 26 Mar 2009 13:42:26 +0000

What security issues are involved with serialized continuations? Is there any way to forge a continuation?

Comment on The State of State on the Web by Untyping » More State on the Web

Untyping » More State on the Web — Thu, 26 Mar 2009 13:35:09 +0000

[...] a followup to The State of State on the Web I want to mention stateless servlets, a relatively new feature of the PLT web server that make [...]

Comment on The State of State on the Web by Noel

Noel — Wed, 25 Mar 2009 17:29:14 +0000

Interesting analogy. Typically another mechanism would be used. A database is the traditional choice. With continuations you might use your language’s inter-thread/process communication mechanisms (in PLT Scheme, this would be channels). Using cookies is certainly viable but not typical.

Comment on The State of State on the Web by apoirier

apoirier — Mon, 23 Mar 2009 17:33:24 +0000

As the lead developer of Nagare, I need to say Nagare is the wrong example to start your post with, because it has the exact same techniques for state management than you use with PLT web: using an history of continuations, saved on the server, is the natural and easier way to code with Nagare. But, then, you can map significative “RESTful” URL on important resources if you want, using CLOS-like generic methodes.

An example on how to map such significatives URL to a Wiki starts at http://www.nagare.org/wiki/wiki7/1 (note how this RESTFul URL points to step #1 of the 7th example of the wiki tutorial).

Also, Nagare comes with several security managers, the form based one using cookies.

That why the complete sentences decribing Nagare are “no *explicit* URL routing” and “No REST *by default*”. But I know the actual lack of document about RESTFul URLs can create miscomprehension.

Else, of course, I totally agree with the rest of your post about the benefits of these different state management policies.

Regards

Comment on The State of State on the Web by dysfunctor

dysfunctor — Sat, 21 Mar 2009 20:56:38 +0000

Thanks Noel. Most informative.

It strikes me that, when you use continuations, the multi-window problem starts to look like the multi-threading problem, with cookies being the data structures shared between threads.

Is this a reasonable analogy?