[PATCH setup 11/11] Use wininet for fetching URLs in direct (non-proxy) case (DO NOT APPLY)

[Jon Turney]

On 03/05/2017 18:08, Ã
ke Rehnman wrote:
> On 2017-05-03 18:37, Jon Turney wrote:
>> Thanks for the patch. So there are a few separate things here:
>>
>> * Pass unknown protocols to wininet
>>
>> This seems a fine idea, but isn't what this patch does.
> Yea, I know, I tested a few different solutions but it was difficult to
> make something clean and proper.

Ok.

>>
>> * Allow wininet to handle file:// URLs
>>
>> I'm a little bit concerned that there may be current uses which rely
>> on the incorrect parsing we do of file:// URLs to work.
>>
>> Otoh, this should fix the file:// URL format we currently mishandle,
>> so is probably worth doing.
> I don't know how that could have ever worked... (or does it?)

Not really, see the other branch of this thread.

>>
>> * What to do with non-URL (i.e. pathname) addresses?
>>
>> Perhaps WinInet can handle these, but assuming it does, is there a
>> good reason to change from using NetIO_File()?
> No. WinInet need proper file://host/path syntax. But the patch I made
> send those unknown protocols to NetIO_File(), right?

Yes.

I'll add a comment that "URLs we can't parse are assumed to be paths" 
and apply this, after I've  released 2.878.

Thanks for working on this.

>> There are also some associated UI issues, in that we give no clue that
>> a pathname is acceptable as a download site, and pathnames and file://
>> URLs are presented terribly in the download site list.
>>
> I saw that. Perhaps fixing that is another patch?

Sure.  I just mention that for completeness, not out of any expectation 
it's going to get fixed :)