[Monotone-devel] Question on layering

[Joel Crisp]

Hi

/unlurks ;-)

I'm toying with the idea of building a small experimental app on top
of monotone. The app has the following characteristics:

a) May need to access practically every entity in a monotone db
b) but, may use a random sliding window on a sub-set of the entities,
possibly with caching
c) is read-only [actually not true, but any writes would be done via
calling monotone for safety]
d) may need to calculate (and cache) additional information relating
to entities in the repository, possibly in a persistent fashion
e) needs a scripting language which can manipulate the underlying C/C++ objects

There are three integration points I could conceivably use:

1) automation [monotone automate]
2) importing the monotone headerfiles and some cpp code and building a
new C++ app which uses the underlying monotone C++ types
3) hit the DB directly

I think from my previous understanding that (1) is the preferred
option, but I am worried about performance, particularly on very large
databases (just one of our apps is 77,000+ files with over 50
developers). Also, (2) gives me lua essentially for free.

Thoughts?

(You may also get monotree updated too as part of this.)

On a side topic, not being able to sync with the monotone master
repository from behind our corporate firewall/proxy is a bit of a
pain. Does anyone have a way to access the venge.net repository over
HTTP tunneling? That and FTP is all our firewall will let through, and
there is no way I will ever be able to change that.

On a second side topic, Redhat Enterprise Linux 4 (Aka Oracle
Unbreakable Linux) appears to only have boost 1.32 - the latest
monotone appears to required boost 1.33 - is this essential?

Thanks

Joel