In a previous episode Markus Buchhorn said...
::
:: At 14:04 11/09/99 +0200, Henrik Nordstrom wrote:
:: >
:: >Also, technically the 'server-side' cache (accelerator) does not need to
:: >be co-located with the actual server. It may be located in another
:: >continent, and have different network connectivity to the origin server
:: >than the end user. So it may both enhance the performance and take load
:: >off the network depending how it is deployed.
::
:: So how is this thing different to a 'mirror' or 'replica' ?
1] mirrors/replicas are complete and never self
updating.. essentially a special case of the surrogate
administrative policy
2] intent
3] the authoritative data is part of the surrogate in some way.. at
least by my definition that views the surrogate proxy and the
traditional HTTP server clustered together as the origin server
:: It seems to
:: come back to 'management' and 'intent' which Joe doesn't like as they're
:: 'unobjective' as a metric.
maybe this needs to be rethought? The taxonomy isn't a protocol
document specifying interoperability, instead it "specifies standard
terminology and the current taxonomy of web replication and caching
infrastructure deployed today. It introduces standard concepts and
protocols usesd today within this application domain"..
concepts can clearly include intent as part of their composition
without providing criteria to evaluate individual deployments against
intent.. to borrow from US Law, several statues require intent as an
element of the crime (1st degree murder for example) while not
providing evaluation criteria for it.. the viewer has to discern from
the circumstances whether or not it fits..
if we call surrogates/mirrors/replicas all the same thing we haven't
fufilled our objective of providing "the current taxonomy of
[wrec]". The technical difference between them does seem to be
1] intent
2] administrative policy
-P
This archive was generated by hypermail 2b29 : Thu Nov 18 2004 - 11:21:27 MST