> But of course, it will allow you to avoid having to scan for
> and parse that header yourself...
>
> Do you have any further thoughts on how we should deal with
> documentation (and
> possibly testing) of what info that can be fetched when?

I would suggest dividing CURLINFO items into two groups, those are derived
from headers, and those that a updated during the transfer of the body.

It would be nice if any info that is derived from a header were available no
later than when libcurl sent the last (empty) header line to the
application, i.e., in the callback that processes the last empty header
line, it would be legal to call curl_easy_getinfo() to retrieve items
derived from the headers.

If that were the nominal rule, then the exceptions to that rule could be
documented. (What I'm saying BTW may be pretty HTTP specific and I'm not
sure how it would apply to other protocols.)

As you mention, that you save you from having to parse the headers yourself,
at least to the extent that libcurl is parsing the headers you are
interested in.
Received on 2007-07-11