15 Jul 2010 lkcl   » (Master)

multitask cooperative single-process http server

egads, mouthful for a subject let alone a title of a wiki page. for various reasons i needed to create a single-process cooperative multitasking HTTP server. it seems, bar one small tolerable bug, that i succeeded in using it to create a special kind of HTTP proxy.

the problem is illustrated by http://gnumed.org being a desktop-based application and using role-based access control on the PostgreSQL database. they want to convert it to a web application, by popular demand, and tried picking random web server frameworks. unfortunately, virtually all web server frameworks assume that the SQL database is to be used for anonymous session cookie management, assume a single (global) username/password and so on.

for gnumed, such assumptions are completely unacceptable: this is confidential patient data, and there's no way that gnumed are going to redesign the database middleware to REMOVE password security.

so i came up with something that will frighten the bejabbers out of any normal web developer: a proxy which is capable of keeping a remote connection permanently open to its upstream target, as long as that upstream HTTP server supports HTTP Connection "keep-alive".

it's… well, i've been yapping about it enough, and so have the gnumed developers: http://lists.gnu.org/archive/html/gnumed-devel/2010-07/msg00326.html

one of the problems with creating such a proxy is that you can't do it with a threaded or a forking architecture: you need to analyse the HTTP stream (check a session cookie and insert one if there isn't one) BEFORE you pass it on to the proxy, but most HTTP proxies are designed to fork first, think later.

the design also has several other uses, one of which is that it solves the issue that pyjamas-web faces. pyjamas-web is the version of pyjamas that actually executes a pyjamas app server-side and translates it into HTML. immediately you can see the problem, there, if you know that HTTP is stateless: how do you re-associate the actual in-memory running python application with the correct browser, when the application is sufficiently large that it is impractical to serialise it to disk?

Syndicated 2010-07-15 20:25:46 from blog/lkcl

Latest blog entries     Older blog entries

New Advogato Features

New HTML Parser: The long-awaited libxml2 based HTML parser code is live. It needs further work but already handles most markup better than the original parser.

Keep up with the latest Advogato features by reading the Advogato status blog.

If you're a C programmer with some spare time, take a look at the mod_virgule project page and help us with one of the tasks on the ToDo list!