Github user spmallette commented on the issue:
https://github.com/apache/tinkerpop/pull/892
> Do you think there is a performance real performance penalty when we do
sorting for serialization?
The graphs you have been working with are small (judging from your
questions on the mailing list you mentioned only a few hundred edges in size).
That's not a big deal - you will likely sort that with no performance penalty.
On the other hand if you were to try to stream out a significantly larger graph
that isn't completely in-memory then I believe the sort would add cost. If you
can demonstrate that there is no cost then I would say we not only normalize by
default but that we deprecate the configuration option and get rid of it. I
think we'd always like to see normalization if there is no penalty for using it.
> I am available for a code review over VOIP if you like to speed up
things, Stephen.
I'm in no rush :smile: - we have no releases officially planned so we can
take our time. Thanks for the offer though.
---
