risc or cisc

[Ian Clatworthy]

Robert Collins wrote:

> I think a mini language is definitely more complex *today* but we're
> adding more and more methods - and the complexity of creating good
> abstractions is getting higher and higher.

I'd really like to encourage you in the strongest possible terms
to experiment with the simple, mini-language approach. I'll be
surprised if it reduces complexity but that isn't the primary issue
IMNSHO. To me, it's all about matching the base technology with
the problem domain. *I* think RPCs are a LAN-appropriate technology
when we have a WAN performance problem to be addressed.

If I can tease you with a stretch goal, if the # of round-trips
(batching aside) for any Bazaar command reaches double figures,
it's a bug. :-)

In fact, I'd expect the number of round-trips to be under 5 for
80% of commands once our network performance is *fully* optimised.
That won't happen tomorrow with a mini-language but a
mini-language makes it possible. And just as importantly, I've
found that having a mini-language as the preferred base technology
puts one into a mindframe ("death to latency!") that makes it
more likely as well.

Of course, YMMV. But I can't tell you how happy it makes me to see
this on the table ...

Ian C.