Outstanding technical issues

[Gearard Woods <gewoods@us.ibm.com>]

Now that a decision has been reached to continue with the current proposal,
I'd like to highlight some technical issues that have surfaced over the
past couple of weeks that should be addressed before the spec can be
considered to be viable.  The most significant of these is the ability to
perform searches which return very large data sets.  With DSMLv2/LDAP, this
is generally accomplished using a paged search control to "page" through
the result set.  The current SPML has removed controls so this is not
currently an option unless, of course, controls are reintroduced and then I
imagine there would need to be some form of schema description for
controls.  Large searches are common in our use cases for performing
reconciliation of target data with the provisioning system's version.
Obviously, if all of the target data must be accumulated in memory before
it can be sent to the client there is a significant scalability problem.

I find it hard to see how it would be accomplished without a major rework,
but without the ability to communicate complex types I think the spec is
just not practical for widespread use.   We have a number of resources that
we provision today where, because of our use of DSML, we are obliged to
encode complex data structures in string attribute values.  This is
obviously problematic for a number of reasons.  My opinion is that
continued effort should be directed at this problem and the related problem
of the schema language, not just for immediate practical reasons but
because this will dictate a lot of the future capabilities of the SPML.

Another issue which has plagued our use of DSMLv2 is in the area of
internationalization.  We touched on this last week, and there are many
approaches to this problem, but DSMLv2 makes it hard and the current SPML
is no better.  It could be argued that internationalization is not the
focus of a spec like this, but the spec should not prohibit the use of
mechanisms that allow the system as a whole to communicate useful
human-readable information.  In DSML, the inability to return multi-lingual
error messages is, I think, a weakness.

Gerry