Complete.Org: Mailing Lists: Archives: gopher: August 2008:
[gopher] Re: Fate of the Protocol (was Re: Gopherness)
Home

[gopher] Re: Fate of the Protocol (was Re: Gopherness)

[Top] [All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index] [Thread Index]
To: gopher@xxxxxxxxxxxx
Subject: [gopher] Re: Fate of the Protocol (was Re: Gopherness)
From: Kyevan <kyevan@xxxxxxxxxxx>
Date: Mon, 04 Aug 2008 17:55:01 -0500
Reply-to: gopher@xxxxxxxxxxxx

Are you offering to host a new list? Not everyone has the resources to 
do so, you know. And of course you wouldn't want to use port 70 to serve 
notgopher, that would be silly. The best idea seems to be to modify the 
Gopher protocol in non-backwards-compatible ways (adding headers of some 
sort to ever resource grab seems to be the big change every other desire 
relies on) and then serve it on a different port, as well as old-style 
Gopher on port 70 - in a similar way to how pygopherd currently serves 
http on port 80 and gopher on port 70.

The real reason to do this is only to try to hit a middle ground between 
http and gopher, really. HTTP is huge, unwieldy, and powerful, while 
gopher is small, simple, and arguably less powerful. Trying to hit a 
balance so we can handle some more complex things (like download 
progress indicators and passing files to the right program on the client 
side) while still keeping things simple enough that cheap, simple (or 
old) hardware is still all you need to get the basic features out of it.

I suppose there's also the gopher+ route of adding extra fields (which 
actually might be a better option), but is there any way to do that when 
you send a magic string? It'd be really nice if a client didn't have to 
know the item type for something before it requested it, but if we keep 
serving on port 70, we obviously have to send that information in such a 
way that it doesn't confuse old clients.



[Prev in Thread] Current Thread [Next in Thread]