[plt-scheme] Planet vs firewalls

From: Jacob Matthews (jacobm at cs.uchicago.edu)
Date: Thu Jan 20 09:50:06 EST 2005

Eli Barzilay wrote:

>But the best solution IMO is close to your last sentence -- it's using
>HTTP as the universal communication protocol -- simply transfer the
>content through a web server.  Operating systems, sysadmins, ISP, etc
>-- all will start helping you instead of complaining about the hole in
>the wall.  (=> for other ports, the default is becoming to be more
>restrictive, and for 80+HTTP you get forwarding hosts, caches, etc
I actually don't have any problem with using HTTP as a protocol for 
requesting and retrieving packages -- that's what it's for, even if 
applications tend to use only a narrow subset of its features. The only 
reason I didn't use HTTP from the outset is that I wanted there to be a 
negotiation phase of the protocol, where recursive dependencies were 
resolved in-protocol before anything was downloaded. I haven't 
implemented that yet, though, and looking back on how people have used 
PLaneT so far that decision now seems like overkill.

When I have time I'll rewrite the server to serve packages as a servlet, 
and let recursive dependencies get handled either with some helper 
servlet that answers recursive-dependency queries, or with the current 
system where requires just get satisfied independently as the compiler 
encounters them.


Posted on the users mailing list.