Note that this is not yet a release announcement. The team is going to release after implementing a bug tracker. The post also outlines recent implementation changes (in the Rust version which is currently being developed).
If you want an overview of the current approach of Pijul, I would recommend Model and Repository layout parts of their documentation (in addition to FAQ).
I’m actually following this project with great interest despite the fact that I suspect the dev community already has all the dvcs’s they want.
The abstract «development community» may have all the DVCS'es «they want», but I know for sure I don’t and I know for sure I am not alone in that (Pijul is not the direction most interesting personally for me, but I hope they do achieve their goal of improving the state of the art around merge).
I’m curious. What do you feel is missing from DVCS’s that you use now? I’m in a similar boat. Git and hg have both given myself a lot of value but there is still a lot in dev workflow that they don’t really completely fill.
I don’t like Git, its model of world is obviously incomplete (reflog is obviously useful, but you can’t either)
First of all I want feature parity with Subversion. I mean, in subversion you can check out a subdirectory and commit in a partial checkout. I agree that in DVCS there has to be some overhead for that, but it should be reasonably possible. (Technically speaking, Monotone allows you to achieve feature parity with SVN here, as in creating a commit remotely without downloading all the files, but I want to be able to make a narrow checkout and commit to it offline, and Monotone’s consistency requirements are non-negotiable: you need to have all the data from all the history to use the local repository).
I want a combination of Monotone’s idea «you can have a single database contain unrelated projects and sync them all by a mask» and generally good idea «you can check what you miss and correctness of all the references» with the idea that «if you have enough partial data to calculate the outcome of the operation, you can do this operation».
I want the DVCS to be able to act as a distributed RDF database if needed (Veracity apparently tried to go in this direction, but by the time it got discontinued I wasn’t sure if it is usable yet).
I also want a DVCS to support incomplete information and synchronization prioritization well enough to be able to replace NNTP…
I really want to try pijul, I am very curious about it. I once tried darcs, and even if I do not stick to it, it have changed the way I am using git. I am eager to try this new tool.