> From ggm@dstc.edu.au Wed Aug 25 19:29 EDT 1999
>
>
> Other point: mirrors usually have a legal and published URL in their own
> right, irrespective of any synthesized mapping from URI-space to their
> instance of the cloned data.
>
> http://mirror.aarnet.edu.au/FreeBSD is a documented published
> URL to reach the same resources as http://www.freebsd.org/
>
> Caches do not exist as terminal objects in themselves (for their content
> that is) in the same way.
>
> I think this distinction matters because in compiling the set of rewrite
> rules for redirection in caches, to supplement any inter-cache ICPlike
> behaviour, you need a complete set of the known URL of the mirrors to
> make a complete set of the possible (and hence derive the preferential)
> rewrite rule to effect.
>
> cheers
> -George
> --
> George Michaelson | DSTC Pty Ltd
> Email: ggm@dstc.edu.au | University of Qld 4072
> Phone: +61 7 3365 4310 | Australia
> Fax: +61 7 3365 4311 | http://www.dstc.edu.au
>
>
>
No-no-no!! You are right that often times mirror do have distinct
URLs. But this is a serious deficiency, and the one that is
disapearing as we speak. among the problems:
* It's not transparent to users; you rely on their good behaviour to
achieve geographical distribution or load balancing.
* If you set up a new mirror to releave the load on the existing site,
at best you can count on NEW users to be load-balanced.
The old URL has been embedded by an unknown number of
documents and bookmark files, and those will not change. They will
continue generating refs to the original site.
* Setting up a mirror is an irrevocable decision. Once you publicize
the mirror URL, you cannot tear it down - people will embed it into
their bookmarks and documents. Yes, you can replace the content on
the mirror with a stub that redirects requests, but fundamentally
you cannot eliminate that mirror URL.
* Most sites prefer to have a single identity, which is undermined by
multiple mirror URLs.
Different products take different approaches to avoiding this
deficiency. But I should get off the soap box; if interested in my
opinions on this issue I discuss this stuff in more details in a paper
"Issues in Web Content Replication", see
http://www.research.att.com/~misha/pubs.html#web.
Regards,
Misha
This archive was generated by hypermail 2b29 : Thu Nov 18 2004 - 11:21:26 MST