Tatsuhiro Tsujikawa <tatsuhiro.t_at_gmail.com> wrote:

> The attached patch 0001 ignores --include option if --metalink is used together.
> I agree that including headers in response is pointless if --metalink
> is used, because it confuses XML parser (if metalink file is
> downloaded from http server) and/or hash check will fail as test2010
> revealed.
>
> The patch 0002 enables test2010 as it is working now.
> I fixed hash value because it seemed wrong.

Great, pushing as I write these lines.

> (Patched your sentences as your last e-mail)
>> And there's something else to ponder. Should option '--dump-header'
>> (-D) work as it is doing right now along with '--metalink' option, or
>> should we make sure that no HTTP headers are dumped nor
>> dump-header file is created when both options are used?
>>
>> If --metalink along with --dump-header is 'fixed', metalink test case
>> definitions will also need fixing.
>
> I feel current behavior is probably OK because it does not break Metalink
> processing and dumping header of downloading remote file is make sense
> to me even in metalink file.

I agree, it is useful to keep it around as long as it doesn't
represent a problem.

While generating test case 2012 (XML file without hash) I've realized
that most likely it would be convenient to further standardise
Metalink messages to further help automatic parsing of these.

I have not fully thought about what a proper format would be, nor all
required uses, I'll try to suggest something next week, unless someone
beats me at it.

Meanwhile I'm pushing 2012 disabled.

Thanks,

-- 
-=[Yang]=-
-------------------------------------------------------------------
List admin: http://cool.haxx.se/list/listinfo/curl-users
FAQ:        http://curl.haxx.se/docs/faq.html
Etiquette:  http://curl.haxx.se/mail/etiquette.html