[jdev] Re: Gaim -> transport

[maqi at jabberstudio.org]

On Sun, 19 Sep 2004, Trejkaz Xaoza wrote:

> This protocol-specific effort is exactly the issue.  When a protocol
> changes, Trillian and GAIM are the _first_ clients to adapt.  By linking
> straight to their protocol plugins, we get the fixes at the same time.
> We don't have to wait for someone to port the fixes to Python.

Well, the discussion is rather old in some way. Take a look at AIM-t: It'd
be no problem to update libfaim and adding server-side buddy support to
AIM-t. However, no one dares to touch AIM-t's C code for quite some time
now (apart from small changes). C ist just no fun.

Apart from this, as a server administrator I really don't want to run a C
transport on my machine. Who knows how many security problems either the
protocol libraries or the Jabber glue code have? For example, as
libfaim/AIM-t neither use the memory pool concept Jabberd uses (or
something similar) nor use some decent string handling library, I bet
there are several severe security problems buried there (as have been in
Gaim's past).

And please, don't rise the "performance issue". For example, C AIM-t
doesn't have a performance issue. It just stops to work properly with many
concurrent connections (as AOL blocks the complete IP if there are too
many unsuccessful logins from one IP which AIM-t doesn't take care for -
that's a software problem no one cared to take a look at!). Plus it has
memory leaks all over the place. With these issues, we don't have to talk
about "performance". ;-)

Regards