There is something to be said about Git and Mercurial. They are clearly the leaders of the pack and there are many good reasons to like them. However there may be as many to hate them too and here’s a short list.
- Linus made it difficult on purpose
- Git is not entirely binary leaving bits in perl
- mercurial is all python; not particularly interesting but package hell
And there is a lot to like about fossil
- single file repo making it easy to backup and restore
- single executable
- based on SQLite; from the SQLite author
- embedded wiki and issue tracker
While there is feature parity fossil features are not embedded in Go and that’s as good a reason to be upset as any. All that said I still want to deploy my latest fossil repo. As such this is the quick-start.
My environment is (a) my MacBook (b) a google compute engine instance running CoreOS. Right now I’m planning to run the remote fossil instance directly on my CoreOS instance, however, it should be moved to a docker container as soon as possible. (should be easy enough).
- install fossil on the target machines. (not going to describe that)
- Let’s start on the mac
- mkdir $HOME/fossil.repo
- cd $HOME/fossil.repo
- fossil init example.fossil
- cd $HOME
- mkdir -p $HOME/src/fossil-scm/example
- cd $HOME/src/fossil-scm/example
- fossil open $HOME/fossil.repo/example.fossil
- echo “richard bucker” >> contributors.txt
- fossil add .
- fossil commit -m “initial import of contributors”
- now let’s register the remote repo
- scp example.fossil <username>@<hostname>:./fossil.repo/.
- fossil remote-url ‘ssh://<username>:<pwd>@<IPADDR>//path_to_repo/fossil.repo/example.fossil?fossil=/path_to_fossil_bin/fossil’
- fossil sync
I was not crazy about the scp in 3.1 but I guess that’s the only way it works. I suppose the intent is to create the repo on the remote system and then clone it locally. These steps and specifying where the fossil executable is located, when using ssh, are just a few more steps. There are clearly a few more steps required in order to get a server running. Maybe another day because it requires setting up a webserver as a reverse proxy.