Loom: Question and/or feature request (or help in establishing a way of working)

[Paul Moore]

One thing I would like to be able to do with the loom plugin is to
switch the order of threads in the loom. I looked at the
documentation, but I can't see anything that looks appropriate.

The reason I want this is that I'm effectively trying to use a loom as
a series of very lightweight branches. I have a copy of the Python
source, and I'm working on a couple of independent changes  I was
hoping to be able to work on them by switching between which is the
first one on the stack, and working on that. Then I can export either
one as a patch as I need to.

I realise that this may seem an odd way to work, however I'm not too
keen on the alternatives. Any one patch may be to only one or two
files, and is quite likely not to even need a rebuild of the
executables (if I'm only changing Python code, for instance).

I keep my pristine mirror in a shared repository under C:\Bzr (this is
so that I can have all my scheduled pulls run in a single location).
However, I do *not* want to work under there - typically I work in
subdirectories of C:\Data.

With that in mind, my alternatives are as follows:

1. Use a full branch under C:\Data. This is hopeless, as a branch
takes 8 minutes to create, which is far too heavyweight for one-file
changes.
2. Use a shared repo under C:\Data. Branches now take 30 sec to
create, but I have an extra directory layer (annoying) and there is
extra time and admin to create the shared repo and keep it updated.
3. Use branches in the main shared repo and then checkouts in the
working directory. Ugh. To create a branch involves switching round
directories. And dropping a branch means doing so in too many places.
Far too much admin.
3a. Use branches in the main shared repo, plus a lightweight checkout
and bzr switch. Still means I'm creating work branches in the
"upstream" repository.
4. Work in the C:\Bzr repository. Well, yes. But I don't want to :-)
(Why should I change the way I work to suit my version control
system?)

And with all of these solutions where I use branches (except (3a)), I
still have to build python.exe in each branch I create. If I reuse a
single branch with different threads in a loom, I can avoid even that.

Am I hopelessly misusing looms? Is there a better way of doing this
that I haven't found yet?

Even an attempt to persuade me that one of the above solutions isn't
as bad as I currently think it is would be useful, as it may tease out
some misconceptions I have.

Thanks,
Paul.