feasability of adding field to sess struct

[Poul-Henning Kamp]

In message <CAJMjcENZJpu-GS8bMNTgjbhmqgB1YpdP_ffwnBcKymTxJZMSJQ at mail.gmail.com>
, John Skopis writes:

>I posted to varnish-misc recently asking how I might be able to
>propgate the http status code from an esi subrequest to the parent
>request, ie esi beresp.status is 5xx, set resp.status to 5xx in
>vcl_deliver.

That is not currently possible, because we deliver the ESI response
as we build it, so by the time the sub request is handled, the status
code is long since sent.

Somewhere on the todo list is "buffered ESI" which would assemble
the entire response in memory, before sending it, and then we can
start to look at such features.

-- 
Poul-Henning Kamp       | UNIX since Zilog Zeus 3.20
phk at FreeBSD.ORG         | TCP/IP since RFC 956
FreeBSD committer       | BSD since 4.3-tahoe    
Never attribute to malice what can adequately be explained by incompetence.