Selective Parsing?

[Rod Widdowson]

One of the things we would like to achieve with parsing RelyinParty.xml is
to separate the metadata provider from the rest of the configuration.  They
really have little to do with each other and so it would be nice if they
were in different services, subject to different reload strategies and so
forth.

In order to do this we really want to be selective about the parsing.  In
particular when parsing the RelyingParty and ProfileConfigurations we do not
want to summon up the metadata provider since that will be slow
(particularly if it involved downloading over HTTP) and might consume
multi-megabytes of memory.

I have been cast around for ways of achieving this.  I have not given up
yet, but I am inviting suggestions on how this might be achieved.  In case
anyone suggests Bean Definition Profiles:  they don't work.  I'll not go
into the stomach churning details, but they work at a level which is too low
for them to be of use to custom schema parsing.

Any suggestions would be more than welcome.  Right now the most promising
direction is to use explicitly named beans for the metadata type beans and
to strip them out  by name before the context is refreshed.  This is not
pretty, it's just the least ugly of the solutions I have come up with so
far.

Thanks

/R