Re: Network stuff

Daniel Vainsencher <[email protected]> wrote:

> Darn, I intended to answer this topic on this thread, but already wrote
> it in the other. In short, I'm for it.
> 
> Anyone objects?
I'm in favour of anything that cleans up the ugliness around filenames,
sockets, streams etc. Now, my understanding is that Flow does a great
deal around this region so perhaps that should be the answer. However,
I also believe (maybe incorrectly) that Flow requires some thread
capability in order to function propelry. Since I can't support threads
on RISC OS and I don't think one can guarantee thread support on
embedded OSs, perhaps I need to stand up for a fallback capability?

Assuming for the moment that Flow is not the best answer for
the immediate future, then these fixups for the older stuff from Mike
are probably a Good Thing. However, reading the comments I get a bit
nervous when I see 'bleeding edge', 'first part of' etc. If Mike is not
claiming that it is ready, perhaps we shouldn't leap too quickly.

I'd really like to see a good answer on all this. Changes to file
handling and sockety stuff are probably some of the most entangled code
around because so many methods make use of them. And _so_ many make poor
use of them....


tim
-- 
Tim Rowledge, [email protected], http://sumeru.stanford.edu/tim
Useful random insult:- Subtle as a well-thrown brick.
From [email protected] Mon May 05 22:26:40 2003
Return-Path: <[email protected]>
Delivered-To: [email protected]
Received: (qmail 30509 invoked from network); 5 May 2003 22:26:40 -0000
Received: from nat1.advantive.com (HELO advantive.com) (204.252.125.74)
  by mail.theinternetone.net with SMTP; 5 May 2003 22:26:40 -0000
Received: (qmail 18411 invoked by uid 0); 5 May 2003 22:26:38 -0000
Received: from unknown (HELO pairhome.net) ([email protected])
  by citrus.advantive.com with SMTP; 5 May 2003 22:26:38 -0000
Message-ID: <[email protected]>
Date: Mon, 05 May 2003 18:27:03 -0400
From: Stephen Pair <[email protected]>
User-Agent: Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US;
	rv:1.3) Gecko/20030312
X-Accept-Language: en-us, en
MIME-Version: 1.0
To: Discussing the Squeak Foundation
	<[email protected]>
Subject: Re: [Squeakfoundation]Network stuff
References: <000101c3133a$580417c0$8ef7fea9@atlantis>
In-Reply-To: <000101c3133a$580417c0$8ef7fea9@atlantis>
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit
X-BeenThere: [email protected]
X-Mailman-Version: 2.1
Precedence: list
Reply-To: Discussing the Squeak Foundation
	<[email protected]>
List-Id: Discussing the Squeak Foundation
 <squeakfoundation.lists.squeakfoundation.org>
List-Unsubscribe: <http://lists.squeakfoundation.org/listinfo/squeakfoundation>,
	<mailto:[email protected]?subject=unsubscribe>
List-Archive: <http://lnx-12.ams-2.theinternetone.net/pipermail/squeakfoundation>
List-Post: <mailto:[email protected]>
List-Help: <mailto:[email protected]?subject=help>
List-Subscribe: <http://lists.squeakfoundation.org/listinfo/squeakfoundation>,
	<mailto:[email protected]?subject=subscribe>
X-List-Received-Date: Mon, 05 May 2003 22:26:40 -0000

Andreas Raab wrote:

>Hi Guys,
>
>It just so happens that I got (once more) bitten by some of the more ugly
>parts of the current networking code and it just so happens that (once more)
>Mike's Network rewrite
>(http://map2.squeakfoundation.org/sm/package/030feb79-34a9-4822-9a6f-d7d2e83
>a0cef) fixes exactly the issues I was struggling with. I don't know if
>that's up on anyone's sleeve but I would really like to see this stuff
>getting into Squeak, mostly because it means that for the stuff I'm doing I
>know I can rely on proper network behavior without changing its fundamental
>semantics.
>
>What are your current thoughts on this issue? The network rewrite has been
>around for a while now and leaving it at SqueakMap probably means it'll rot
>and die there with anyone having similar problems at hand reinventing the
>solutions he's already bundled up.
>

Sounds good to me.  Also, I would like to get the aspects of Comanche 
that touch on this merged with Mike's stuff (like SocketStream).  I've 
recently spent some time and made a lot of progress in supporting 
HTTP/1.1 (persistent connections now work! as well as OPTIONS, TRACE, 
etc...webdav is coming as well).  But that's all server side stuff.

It doesn't look like the Mike's stuff does much to HTTPRequest or 
HTTPSocket.  Am I wrong?  Is there an HttpClient?  I would like to merge 
those with Comanche's versions (but it can wait a while).  My thoughts 
are that HttpRequest/HttpResponse should take a minimalist 
approach...focusing only on reading and writing requests and response 
onto streams.  Comanche adds a bunch extra stuff to HttpRequest and 
HttpResponse that really belong elsewhere (and I'm minimizing them with 
each release of Comanche).  Comanche's versions of 
HttpRequest/HttpResponse have been more focused on the server side (so 
writing a request, and reading a response are not well tested)...so, if 
someone might like to work on getting Comanche's HttpRequest and 
HttpResponse to work well on the client side, and to write an 
HttpClient...I'd love the help.

- Stephen