Jason Andrade wrote:
> On Sun, 5 Nov 2000, Mark Day wrote:
>
> > > Apologies if this is an ignorant question, but where do mirror sites
fit
> > > in the scheme of things.. are they just less sophisticated
> > > reverse proxies?
> >
> > Apologies in turn if I now ask an ignorant question, but what aspect of
> > mirror sites would you want to standardize? Answering that question in
a
> > little more detail may make it easier to determine where it should be
> > discussed.
>
> i look after a couple of mirror sites and also look after some mirror
> distribution networks for archives. there's little to nothing in the
> way of standards for mirrors. there are a number of tools available and
> a loose `confederation' of mirror admins who exchange knowledge but that's
> about it. the wheel often gets re-invented in new and wonderful ways.
>
> from what i've seen WREC/WEBI seem to be able producing standards and
> it would have been useful to know if this could be applied to mirrors
> and if it clearly couldn't maybe there needs to be another forum for
> those sorts of discussions/work.
>
> apart from "raw" mirror sites, there's also been some work on redirect
> protocols being used into mirror sites and things like that are not
> well understood or deployed yet, which they might be if there was a
> clear way to implement them (or not as the case may be)
>
> as an example, not too many squid cache admins allow for close to gigabyte
> sized individual files in their archive, yet mirrors are sending terabytes
> of this kind of data around the net daily.
As Jason points out, mirror servers have the capability to create surrogates
for services which are not easily replicated using current caching and CDN
technology. Of particular interest to me is uncachable content generated by
code executing on the mirror.
Currently, creating a mirror site which includes such executable content is
very difficult when the mirror is to be installed on a server that is not
identical to the origin server. This is because of the absence of standards
for the operating environment, language and API for executable code and the
mechanisms by which such code is transmitted to the mirror server. This
problem is very closely related to the problem of establishing standards for
installing new services on an Extensible Proxy.
Current CDN operators and ASPs seem happy to define proprietary mechanisms
for creating mirrors of executable content, and will require that their
customers conform to their proprietary APIs in order to take advantage of
them. This is seen as a way of locking their customers into their
particular server networks. As these APIs become more complex,
interoperability will become impossible. Definition of some basic standards
would at least allow some executable content (using only basic services that
can be standardized) to operate across CDN and ASP server networks.
Micah Beck
This archive was generated by hypermail 2b29 : Thu Nov 18 2004 - 11:21:29 MST