Daniel Stenberg wrote:
> On Thu, 13 Sep 2007, Michal Marek wrote:
>> I've a (half-working :)) patch to automatically apped ;type=<a|i> in
>> such cases if not already present, to make it consistent with the
>> non-proxy case. Do you think curl should behave that way and I should
>> finish the patch, or does the user need to take this difference into
>> acount?
>
> I think it sounds like a good idea, since when speaking FTP over a HTTP
> proxy I think the proxy needs as much extra hints as possible to make an
> as good transfer as possible!

OK, here is the patch, I had to adjust test79 and test208, other tests
passed:
TESTDONE: 345 tests out of 345 reported OK: 100%
TESTDONE: 388 tests were considered during 156 seconds.
TESTINFO: 43 tests were skipped due to these restraints:
TESTINFO: "curl lacks any scp support" 4 times (601, 603, 605, 607)
TESTINFO: "no stunnel" 18 times (300, 301, 302, 303, 304, 305, 306, 307,
308, 400, 401, 403, 404, 406, 407, 408, 409, 509)
TESTINFO: "Resolving 'ip6-localhost' didn't work" 1 times (241)
TESTINFO: "curl lacks any sftp support" 12 times (600, 602, 604, 606,
608, 609, 610, 611, 612, 613, 614, 615)
TESTINFO: "configured as DISABLED" 1 times (351)
TESTINFO: "curl lacks idn support" 1 times (165)
TESTINFO: "curl lacks netrc_debug support" 6 times (130, 131, 132, 133,
134, 257)

Is patching Curl_http() the correct approach, or should this be done
somehow earlier? Also, I didn't find a better way to find out the
original url protocol than string-comparing the url.

Michal