Re: [draft-nottingham-http-link-header-06] concerns about Link header

On Mon, Aug 31, 2009 at 2:19 PM, Noah Slater <nslater@tumbolia.org> wrote:

> On Mon, Aug 31, 2009 at 02:09:18PM +0200, Sam Johnston wrote:
> > Not only is linkrot all but eliminated but the links are higher
> quality/less
> > opaque (eg http://microsoft.com/win7 rather than
> http://tinyurl.com/decafbad)
> > and therefore more secure, plus as everyone's using the same one things
> like
> > visited link colouring still works.
>
> These "short" links increase link rot. Now instead of relying on a single
> domain
> name, and a single host, you are relying on two of each. And as these
> linking
> services go off-line, entire swaths of the Web break silently.
>

This is OT but worth clarifying nonetheless.

Assuming we agree there's a need for short links[1] then I recommend people
use their existing [generally already as short as possible] domain name
rather than some other domain that's likely to be forgotten, expire, break,
etc. Using the same domain is not only more transparent for users (consider
wordpress.com vs wp.me) but can be significantly more performant. URL
shorteners currently add hundreds of milliseconds to requests higher level
proposals like Adjix<http://www.scripting.com/stories/2009/08/19/howToFixUrlshorteners.html>are
almost certainly even more expensive. Rather than having to
recursively
resolve two domains (possibly including CNAMEs), complete two TCP handshakes
and then conduct two separate HTTP requests, the redirect overhead can be
reduced to negligent levels by reusing a single HTTP connection.

Those sites that still choose to use a "domain hack" (like
wordpress.com->wp.me, techcrunch.com->tcrn.ch) should advertise both and let
clients choose - they can go for the shortest, most descriptive (longest) or
most performant (same domain).

As for linking services going offline, it's much the same as the situation
with third-party DNS services <http://cr.yp.to/djbdns/third-party.html> - if
the resource is down there's no point having the pointers to it. The best
situation is where they are colocated, for example being generated by a CMS
[plugin] (e.g. Drupal, Wordpress) - that way anything which takes out the
short links will also take out the target content and when the content is
taken offline the links are cleaned up too.

I've thought about this problem for some time now and have been unable to
come up with a better option, but if you can then I'm all ears as it's not
too late to change tack. If we don't do something then others
will<http://www.mnot.net/blog/2009/04/14/rev_canonical_bad>
.

Sam

1. Even if Twitter were rich text and thus hid the links we'd still have SMS
and physical applications to think about like radio, print and TV
advertising. Conversely, canonical links want to be as descriptive as
possible so as to contain lots of SEO juice. That is, the requirements are
diametrically opposed.