From tjg at pentaside.org  Tue Dec 16 16:27:35 2008
From: tjg at pentaside.org (Tayssir John Gabbour)
Date: Tue, 16 Dec 2008 17:27:35 +0100
Subject: [drakma-devel] Surprising jump in time when sending binary
Message-ID: <4947D6F7.1090602@pentaside.org>

Hi!

If I send a certain amount of bytes (20344), it takes .3 seconds. But if 
I send one that's just a single byte bigger than that, it takes over 5 
seconds.

Aside from that unexpected jump, it otherwise looks linear with the 
number of bytes sent.

I hunted down the problem, and discovered that if I slept for .1 seconds 
before drakma:HTTP-REQUEST calls FINISH-OUTPUT, the problem goes away. 
(Until, I suppose, I send something much bigger.)

Any ideas? Maybe this is a Lispworks problem?

(Lispworks 5.0.0, WinXP, Drakma 0.11.5)


Thanks,
Tayssir



PS: My test --

   cl-user> (let* ((num-octets 20345)
                   (octets (coerce (loop repeat num-octets
                                         collect (code-char 50))
                                   'vector)))
              (time (drakma:http-request "http://..."
                                         :content-length num-octets
                                         :content-type 
"application/octet-stream"
                                         :want-stream nil
                                         :method :post
                                         :content octets)))

User time    =        0.000
System time  =        0.010
Elapsed time =        5.225
Allocation   = 37612 bytes

-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: drakma-diff.txt
URL: <https://mailman.common-lisp.net/pipermail/drakma-devel/attachments/20081216/51dbd7be/attachment.txt>

From tjg at pentaside.org  Wed Dec 17 10:51:51 2008
From: tjg at pentaside.org (Tayssir John Gabbour)
Date: Wed, 17 Dec 2008 11:51:51 +0100
Subject: [drakma-devel] Surprising jump in time when sending binary
In-Reply-To: <4947D6F7.1090602@pentaside.org>
References: <4947D6F7.1090602@pentaside.org>
Message-ID: <4948D9C7.9080907@pentaside.org>

Hmm, let me actually test that on the newly-released LispWorks before 
anyone spends time on this...

Tayssir


Tayssir John Gabbour wrote:
> Hi!
> 
> If I send a certain amount of bytes (20344), it takes .3 seconds. But if 
> I send one that's just a single byte bigger than that, it takes over 5 
> seconds.
> 
> Aside from that unexpected jump, it otherwise looks linear with the 
> number of bytes sent.
> 
> I hunted down the problem, and discovered that if I slept for .1 seconds 
> before drakma:HTTP-REQUEST calls FINISH-OUTPUT, the problem goes away. 
> (Until, I suppose, I send something much bigger.)
> 
> Any ideas? Maybe this is a Lispworks problem?
> 
> (Lispworks 5.0.0, WinXP, Drakma 0.11.5)
> 
> 
> Thanks,
> Tayssir
> 
> 
> 
> PS: My test --
> 
>   cl-user> (let* ((num-octets 20345)
>                   (octets (coerce (loop repeat num-octets
>                                         collect (code-char 50))
>                                   'vector)))
>              (time (drakma:http-request "http://..."
>                                         :content-length num-octets
>                                         :content-type 
> "application/octet-stream"
>                                         :want-stream nil
>                                         :method :post
>                                         :content octets)))
> 
> User time    =        0.000
> System time  =        0.010
> Elapsed time =        5.225
> Allocation   = 37612 bytes
> 
> 
> ------------------------------------------------------------------------
> 
> _______________________________________________
> drakma-devel mailing list
> drakma-devel at common-lisp.net
> http://common-lisp.net/cgi-bin/mailman/listinfo/drakma-devel



From edi at agharta.de  Wed Dec 17 11:14:56 2008
From: edi at agharta.de (Edi Weitz)
Date: Wed, 17 Dec 2008 12:14:56 +0100
Subject: [drakma-devel] Surprising jump in time when sending binary
In-Reply-To: <4948D9C7.9080907@pentaside.org> (Tayssir John Gabbour's
	message of "Wed, 17 Dec 2008 11:51:51 +0100")
References: <4947D6F7.1090602@pentaside.org> <4948D9C7.9080907@pentaside.org>
Message-ID: <uljuf59zj.fsf@agharta.de>

On Wed, 17 Dec 2008 11:51:51 +0100, Tayssir John Gabbour <tjg at pentaside.org> wrote:

> Hmm, let me actually test that on the newly-released LispWorks
> before anyone spends time on this...

Yes, and please let us know the results.  This sounds strange, but
right now I don't have the time to look into this.

Thanks,
Edi.