[Gzz] ARGH!

[Benja Fallenstein]

Here's another problem with saving I hadn't considered yet. Our saving 
process relies on being able to create diffs between SliceVersions. To 
do that, we need to be able to put the contents of two slice versions in 
a HashSet so as to compute the difference.

This needs to work across saving/loading: It must be possible to load 
the last saved SliceVersion, and compare it to the current SliceVersion 
for diffing.

Therefore, spans need to be comparable and hashable. (Not that you 
wouldn't expect them to be anyway, but this is why they HAVE to be.)

You can hash a span by XORing its bounds with the hash of its scroll block.

You can hash a scrollblock by hashing its ID; you can compare  two 
scrollblocks by comparing their IDs.

However, you cannot hash and compare a non-finalized scrollblock that way.

You  could settle for comparing TransientTextScrolls by object identity, 
but then a *saved* TransientTextScroll would not be equal to a 
PermanentTextScroll with the same id.

The only way I see around this is comparing TTSs by object identity, and 
PTSs by ID, and making sure that there is no PTS with the same ID as an 
existing TTS. The problem is: there is no way to make *sure*. I now let 
the TTS put themselves into ScrollBlockManager's msCache (UGLY), but 
that only works when you create PTSs through ScrollBlockManager, not 
directly.

ARGH!

- Benja