On Tue, Apr 22, 2008 at 12:46 AM, Paul Gilmartin <[EMAIL PROTECTED]> wrote:

>  I thought I had done the CRLF right.  Now I need to look at
>  my own code to verify.  In any case, that sort of coding
>  error should be data-sensitive and reproducible, whereas the
>  failure I'm experiencing is intermittent.

My REXX is weak and I could not grasp all the fine details in your
code. I think you leave the loop when the received buffer starts with
a LF. Since you should not have any expectations about how lines span
the network buffers or the records you get from tcpclient, that looks
fishy to me.

>  But it's possible that in its lifetime it has never encountered
>  CRLFCRLF, which I assume would be needed to trigger the failure
>  you envision.
>
>  I'm looking at WGET.  Is it proxy-able?  And I see:
>
>    '| h: httpsplit',                  /* Split header and body       */
>    '| xlate from 850 to 1047',        /* ASCII to EBCDIC             */

No. The HTTPSPLIT stage separates the header and the body, so the
xlate only deals with the header (that I want EBCDIC). The body of the
transfer is passed to the WGET REXX which interprets the mime headers
to make up its mind.

PS Earlier this week I was thinking about a change to make it use a
conditional GET (to download only if the file on the web is more
recent than my copy). But did not have time yet.

Rob
--
Rob van der Heij
Velocity Software GmbH
http://velocitysoftware.com/

Reply via email to