W3C home > Mailing lists > Public > xproc-dev@w3.org > December 2008

Re: Calabash and http-request

From: David A. Lee <dlee@calldei.com>
Date: Tue, 23 Dec 2008 08:04:16 -0500
Message-ID: <44351913FD0149508F572458350C4356@calldei.com>
To: "Norman Walsh" <ndw@nwalsh.com>, "XProc Dev" <xproc-dev@w3.org>

--- Quoteth Norm

I see where you're going, but I think the WG has consensus that the
p:http-request step should be a fairly rich step. I also think there's
a natural answer to the questions you pose: the step should be media
type agnostic. So cookies are in, Java, ActiveX Controls, etc. are
out.
----

Would this then  become a requirement ? or a vendor implementation decision 
?
Cookies are particularly problematic because to fully implement them 
requires state.
That is they may have an expiration way into the future (1 year is not 
uncommon),
so you'd need to keep a cookie cache around and also document the side 
effects.
http-request would then become stateful.   Its functionality could depend on 
previous calls
either in the same pipeline, same xproc "session" or previous 
incarnations/runs of the system.
Then you'll need to expose a way to "clean" the cookie cache, or parts of 
it.

I personally want to stay way away from that slipperly slop.  There's 
existing tools out there
that handle this and they are non-trivial.  Any trivial implementation will 
fail in non-deterministic or
implmentation-dependant ways,  which is worse, IMHO, the not handling 
cookies at all.
If this becomes a spec requirement then you've got a lot of paragraphs to 
write.  You'll need to
define what subset (if any) of the Cookie specs are required, what are 
optional and what the
behaviour would be if the optional parts are omitted.


-David
-----------------------------------------------------------
David A. Lee
dlee@calldei.com
http://www.calldei.com
Received on Tuesday, 23 December 2008 13:05:00 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Tuesday, 23 December 2008 13:05:00 GMT