Others have pointed out that I should stick to
RFC 1918 address space. But again, this is a
lab network and to use the words of another,
one of the things I want to do is make it much
easier to "parse visually" my route tables.
Think of it as a "metric system" type of numbering
plan. The 1 and 100 nets would not be advertised
via BGP obviously...not a hijack situation at all.
If I take into account the possibility that this
lab will have later requirements to connect to
the internet, all I have to do is have a NAT plan
in place...one that even takes into account that
the 1 and 100 nets could become available some
day, correct?
Thanks to those who have responded so far.
Might want to use networks 4/8, 16/8, and 64/8 - they stand out
nicely when looking at net numbers in hex or binary. 
Brennan,
If you want your routes to be human parse'able, I recommend running your lab in full IPv6 mode. That way you take Valdis's recommendation to a whole new level (and base number system).
Plus... Whats the point of having a lab that only uses 1982/1983 addressing techniques (1/8, 10/8, 100/8 are classfull addresses). Labs are meant to push the limits of todays technology and experiment with future concepts. IPv6 matches that criteria.
Martin
If your net 1 and your net 100 talk to each other in your lab, what sort
of NAT plan would allow your net 1 to distinguish between your net 100
and the real net 100?
Really... There are three different zones of RFC-1918 space, so your routing
tables should still be pretty easy to visually parse.
Owen