On 31/10/2013 21:04, Allan Jude wrote:
> I wonder if the http+pkg:// protocol can solve this, likely will require
> a patch to fetch to implement the logic to do the dns lookup and make
> the proxies request for the real hostname

It's pkg+http:// or pkg+https:// or pkg+ssh:// or -- well, you get the idea.

pkg+http:// is really exactly the same as the current http://
PACKAGESITE URLs -- the new code pretty much checks that the first four
characters are 'pkg+', moves the string pointer to the following
character (ie the h in http://) and then carries on exactly as it works
right now.  We'll be accepting either form certainly throughout the
lifetime of 1.2.x release, but printing a warning to switch to
pkg+http:// where relevant.

The reason for doing this is that according to RFC 2616 in a URL of the
form http://some.add.ress/ the 'some.add.ress' bit has to be either an A
or a CNAME record that can be resolved to an IP address.  Since we can't
change the meaning of 'http://' in URLs, we've just invented our own URL
scheme.  Once it is in reasonably widespread use it's pretty much going
to be de-facto accepted, and we can apply to ICANN to have the scheme
officially registered.



Dr Matthew J Seaman MA, D.Phil.
PGP: http://www.infracaninophile.co.uk/pgpkey

Attachment: signature.asc
Description: OpenPGP digital signature

Reply via email to