[Gzz] 1st, 2nd (Benja)

[Benja Fallenstein]

- Make interpolation work better: only interpolate when needed; 
interpolate for 2.5 and not 10 seconds. It's surprising just *HOW MUCH* 
better Gzz feels after the latter change (a single constant).

- Working on serialization. Here's my plan for solving the difficulties 
I described in my earlier mails:

a) Serialization of Spans works only through MediaOutputStream and 
MediaInputStream, which extend Object*Stream. Only additional thing they 
provide is a getMediaserver() method. This solves two things at once: It 
provides a context for saving and loading the scrolls, and it indicates 
a policy, namely that data may be saved/loaded during de/serialization. 
Attempting to serialize with Object*Streams will throw 
NotSerializableExceptions. This makes me feel more at ease with the 
whole concept, too, because by using Media*Stream, the client agrees to 
the assumptions behind it.

b) Unify PermanentTextScroll's TextSpan and TransientTextScroll's 
TextSpan into a single class (that has a (transient) pointer to a 
TextScroll). That way, we can serialize it pointing to a TransientTS and 
deserialize it pointing to a PermanentTS.

- Thinking yet more about vobs.

- Thinking about structure of our work-- I'll post some more detailed 
notes in a few days about this, but basically: Divide our work in a 
number of 'profiles' (plain zzstructure, fancy text editing, distributed 
work...) which require different kinds of research, so that we know what 
we're working on and how stable that is. Aim for a 1.0 release of the 
vanilla zzstructure implementation, which is more user-friendly, has 
clean code and some stable time for fixing bugs and thwarts. Stop the 
thinking that there are a million things to do before we're 
feature-complete and can start thinking about real usability... there 
are a million things to do, but they can go in later versions.

I will be on vacation for the next three weeks, so I'll process email 
and cvs in batch, about 1-2 times a day I think.

-b.