InCommon metadata support

[Tom Scavo <>]

On Tue, Feb 16, 2016 at 4:10 PM, Cantor, Scott 
<>
 wrote:
>
>> With the MetadataFilter in place, it takes 90-120 seconds.  So something 
>> with
>> the signature process is chewing up a lot of time.   I’d probably suspect
>> something in the XML canonicalization that happens as part of signature
>> validation.  It seemed to be doing a ton of malloc calls, so I suspect that
>> there’s something in the XML canonicalization that does that, and it 
>> increases
>> exponentially with the size of the metadata file being validated.
>
> Yes, it does, that's more or less known about that code (it's not mine, I'm 
> just stuck as the only maintainer left). I had thought the DOM portion 
> alone was nearly as bad, but apparently not. Most of the performance issues 
> with it had been reported on even larger files, but it's not that 
> surprising.

I need to complicate the situation by describing a change that was
made to InCommon metadata to save space. (Ian, feel free to jump in
with details.) We took this opportunity to optimize the namespace
declarations in InCommon metadata. Essentially we elevated namespace
declarations to parent elements wherever possible. In some cases,
those declarations were elevated all the way up to the root element.
This saved a lot of space but perhaps it is contributing to the
behavior Jeffrey is seeing, I don't know.

One thing we could do, I suppose, is generate an aggregate on MDQ-beta
without the namespace optimization and see if that matters. Jeffrey,
if we did that, would you be willing to test against that new
aggregate? (It would have the same content, just different namespace
declarations.)

Tom