[TAP] sync vs async
paul.harrison at manchester.ac.uk
Tue Mar 10 02:42:43 PDT 2009
On 2009-03 -09, at 18:01, Roy Williams wrote:
> Thank you Doug, for this insightful and constructive message.
> I for one believe in the open systems approach that you describe,
> the idea of a toolkit of services that can be chained together, the
> idea of building our system gradually and with flexibility.
I am not sure who in the VO *doesn't* believe in open systems - it is
pretty much the guiding principle by which we are all working.
> As for the other way, well, twice in my past life I have been
> involved with a 'big system' that has failed. Each time it was that
> same reasons: failure to involve customers, inadequate use cases,
> too much system for them to learn, emphasizing capability over
> robustness, and such a late delivery that customers had already
> built alternative ways to get their work done.
However, to play Devil's advocate, the biggest success story in the
general arena of astronomical data delivery systems in the last few
years has been Google Sky - a closed system with a dedicated server
and client - the very definition of a 'big system'.
> As Doug points out, our customers have an alternative to the VO: we
> are like Travelocity/Expedia, an integrator of data sources, we are
> not a telescope, which original data. So we need to be attractive to
> customers. Starting now.
Again we all agree, however we need to recognise the conflict between
making the software user friendly and the protocols too simplistic. We
also must remember that the Data Centres are not the "users" - the
users are the individual astronomers in their university departments.
We often hear the argument that we must not make something too complex
because data centres will not implement it, but unless the data
centres go through a little pain to take on most of the burden of
complexity, then that is passed on to the end users.
More information about the dal