From edi at agharta.de  Sun Nov  5 23:48:43 2006
From: edi at agharta.de (Edi Weitz)
Date: Mon, 06 Nov 2006 00:48:43 +0100
Subject: [flexi-streams-devel] New version 0.7.2
Message-ID: <uwt69ihes.fsf@agharta.de>

ChangeLog:

  Version 0.7.2
  2006-11-06
  Removed unnecessary CHECK-EOF-NO-HANG also for in-memory streams (see 0.5.8)

Download:

  http://weitz.de/files/flexi-streams.tar.gz

Cheers,
Edi.


From vodonosov at mail.ru  Tue Nov 14 18:34:07 2006
From: vodonosov at mail.ru (Anton Vodonosov)
Date: Tue, 14 Nov 2006 20:34:07 +0200
Subject: [flexi-streams-devel] recovering from flexi-stream-encoding-error
Message-ID: <FBENJDGGPLPHJJCNFLMJEEDNCDAA.vodonosov@mail.ru>

Hi!

It sometimes necessary to recover from 
flexi-stream-encoding-error.

For example, it's inconvenient now
to use drakma for sites with broken
utf encoding (like this livejournal
page http://hzweiblum.livejournal.com/)

I can see this page in my browser (with
? instead of badly encoded chars), but
(http-request "http://hzweiblum.livejournal.com/")
fails signaling flexi-stream-encoding-error.

It will be good to have possibility to 
supply a character that will be used 
as a substitution for invalid byte sequence.

use-value restart is added to the 
get-char-code function in the attached
patch.

Regards,
-Anton
-------------- next part --------------
A non-text attachment was scrubbed...
Name: flexi.diff
Type: application/octet-stream
Size: 2818 bytes
Desc: not available
URL: <https://mailman.common-lisp.net/pipermail/flexi-streams-devel/attachments/20061114/cd089d1b/attachment.obj>

From edi at agharta.de  Tue Nov 14 20:56:28 2006
From: edi at agharta.de (Edi Weitz)
Date: Tue, 14 Nov 2006 21:56:28 +0100
Subject: [flexi-streams-devel] New version 0.8.0 (Was: recovering from
 flexi-stream-encoding-error)
In-Reply-To: <FBENJDGGPLPHJJCNFLMJEEDNCDAA.vodonosov@mail.ru> (Anton
	Vodonosov's message of "Tue, 14 Nov 2006 20:34:07 +0200")
References: <FBENJDGGPLPHJJCNFLMJEEDNCDAA.vodonosov@mail.ru>
Message-ID: <uejs5u4qr.fsf@agharta.de>

On Tue, 14 Nov 2006 20:34:07 +0200, "Anton Vodonosov" <vodonosov at mail.ru> wrote:

> It sometimes necessary to recover from flexi-stream-encoding-error.
>
> For example, it's inconvenient now to use drakma for sites with
> broken utf encoding (like this livejournal page
> http://hzweiblum.livejournal.com/)
>
> I can see this page in my browser (with ? instead of badly encoded
> chars), but (http-request "http://hzweiblum.livejournal.com/") fails
> signaling flexi-stream-encoding-error.

Hmm, works for me.

> It will be good to have possibility to supply a character that will
> be used as a substitution for invalid byte sequence.
>
> use-value restart is added to the get-char-code function in the
> attached patch.

I've added your code to the new release.  See also the new special
variable *SUBSTITUTION-CHAR*.

Thanks,
Edi.


From vodonosov at mail.ru  Tue Nov 14 22:01:29 2006
From: vodonosov at mail.ru (Anton Vodonosov)
Date: Wed, 15 Nov 2006 00:01:29 +0200
Subject: [flexi-streams-devel] RE: New version 0.8.0 (Was: recovering from
	flexi-stream-encoding-error)
In-Reply-To: <uejs5u4qr.fsf@agharta.de>
Message-ID: <FBENJDGGPLPHJJCNFLMJAEDPCDAA.vodonosov@mail.ru>

>> I can see this page in my browser (with ? instead of badly encoded
>> chars), but (http-request "http://hzweiblum.livejournal.com/") fails
>> signaling flexi-stream-encoding-error.

>Hmm, works for me.

Yes, I've muddled they up. This signals a flexi-stream-encoding-error:
 (http-request "http://www.livejournal.com/interests.bml?int=tao&page=9")

Tried new version, all works fine.

Thanks,
-Anton




From edi at agharta.de  Mon Nov 20 22:32:50 2006
From: edi at agharta.de (Edi Weitz)
Date: Mon, 20 Nov 2006 23:32:50 +0100
Subject: [flexi-streams-devel] Re: [hunchentoot-devel] serve-event support
	for SBCL and CLISP
In-Reply-To: <391f79580611201350u480730e8q60f0a838202cc0c3@mail.gmail.com> 
	=?iso-8859-1?q?=28Lu=EDs?= Oliveira's message of "Mon,
	20 Nov 2006 21:50:16 +0000")
References: <391f79580611191325obc1ad00y831ae7e6724bec7c@mail.gmail.com>
	<u64d9izrs.fsf@agharta.de>
	<391f79580611201350u480730e8q60f0a838202cc0c3@mail.gmail.com>
Message-ID: <uwt5phhpp.fsf@agharta.de>

[Cc to flexi-streams-devel]

On Mon, 20 Nov 2006 21:50:16 +0000, "Lu?s Oliveira" <luismbo at gmail.com> wrote:

> Hmm, are you supposed to be able to write a sequence of octets onto
> a flexi-stream with element-type character? I can't tell from the
> documentation, though it apparently it works on most Lisps but
> CLISP. If so, what exactly is the element-type for?

Several Lisps have "bivalent" streams where you can write octets as
well as characters to the same stream regardless of its element type.
(I'd say that in this case the element type is just some kind of
"hint" provided by the person who sets it.)  If I were to decide,
flexi streams would also be bivalent - FLEXI-STREAMS is flexible
enough to support that (note the superclasses of FLEXI-INPUT-STREAM,
for example), but apparently CLISP isn't.

I'm not sure if anything in the ANSI spec (or in the Gray streams
specification) explicitely forbids bivalent streams.  If that's the
case, at least LispWorks, AllegroCL, and SBCL are non-confirming, but
I like their behaviour.  My personal opinion is that CLISP's Gray
stream implementation is simply suboptimal.