
Jon Smirl wrote:
Six people have various modifications to u-boot hosted on github. These projects aren't linked to each other.
I just talked to the github people. To fix this the main u-boot repo needs to be pushing a clone of itself to github. This is free to do, just make a git hub account and then set your repo to mirror changes there. Once the mirror is in place, github users can fork from from it. Now github can links these forks to the root repo and not keep six copies.
The linux kernel git tree is already being mirrored at github.
The effect of this is to create a public place where people can work on patches for u-boot.
Hi Jon,
This seems like a good idea to me but bears thinking about...
Just to reiterate some history, U-Boot was hosted on SourceForge for a long time, but SF became slower and slower. When it became intolerably slow, Wolfgang took the bits off and we transitioned to git, hosted on denx.de. This has worked *extremely* well. Even for people that are forking and not pushing (all of) their patches back, git has to be a HUGE win over trying to to the same thing with SVN. (At CIdeas we use to clone the SVN repository and then control local changes with RCS - bleah!)
It looks like github's business model is reminiscent of SF (and borrows from BitMover/BitKeeper too - pay to be private). It appears to be a lot less grandiose that SF - only doing git repo hosting, not the whole development lifecycle model (repo, bugtracking, web pages, etc.).
On the plus side ---------------- * It costs denx.de nothing to mirror the master to github
* It spreads the load (although denx.de seems to be responsive to date)
* Since git is *distributed*, github is just another repo and so we aren't "migrating onto" it and, if their business model fails, we wouldn't have to "migrate off" of it.
* It would encourage more public "private" repos - currently there are a lot of repos that are private or are publicly available but not advertised / not discoverable. This could be a Good Thing for cross pollination and getting wider testing and acceptance of patches before they get included into the mainline. Or not.
* Wolfgang already has the denx.de infrastructure set up, but this may give denx.de relief on the sysadmin work.
On the negative side -------------------- * Wolfgang would potentially give up some (mostly illusionary) control. * Brand dilution? Would people get confused which was master? Do we care?
Questions --------- * How stable is github? What is their long term viability? * Do we care? * Who is github? What is their relationship with EngineYard? * (EngineYard is pretty expensive host for a free service.) * (http://logicalawesome.com/ are the guys behind github.) * Do we care? * What questions haven't I asked?
Thanks, gvb