On 03/19/2015 04:59 PM, Mega wrote: > Hi Yves > > Am 18.03.2015 um 12:45 schrieb Yves Blusseau: >> Using git to store the packages and using branch to group the packages by >> version/architecture is really a good thing >> and it's not dependent to how and where the files are host. >> But using http to get the packages is a bad thing for me because is >> dependent of how the http server expose the >> files/trees. > > There is no git on the client machine,so we can't use git to fetch the > packages. > Using git on the client has advantages. You can use git with locally repos or with remote repos. You can keep tab on the client and have a version controlled backup You could let the client pull files that you stage remote. You could poll the client to a branch, diff config files against a release branch, merge to an update branch and poll/push to the client.
Mayby even intergrate some contentious delivery tool and have virtbox, qemu, qemu/gns3 as targets What does it take to have git on the client? A little history background. Maybe we should look at that the world of 2004 has changed alot. Then I was lucky to get a ADSL connection that didn't work due to long copper lines and to many connection points. We did build our own community network combining RadioAccess to a fiber. Now I have my own fiber access to the house and the Firewall has to be upgraded to handle much more. I'm still using a i686 pro as fw and its good for 70-80 Mbit/s and is to slow for running a VPN Today we have higher speed in our networks and we have multiple wans and VPN In sweden I can get a 4G dataplan for 8 EUR a month it's limited (1000/128 kbit)on transmission rate, unlimited in volume perfect as a mgt or backup link. I don't want my data in a cloud but my "owncloud" as my data is important to me and want offsite backups. My old DDS2(2-4 GB/tape tape station doesn't cope anymore as I have x TB of data. I have been testing a FreeNas on an old HW it's nice, I would go for btrfs on centos and containers for services. Transmission speed is about handling a number of IRQ/s and CPU to handle packet processing. If somone builds an add-on board to a RPi 2 with POE eth interface and AES NI support Building the board stackable and with a switch/bridge module. The RPi2 as an mgt interface and use SPI to talk with the switch /router using some form of structable text(Yaml). Or just stack several RPi2 together. >> For me it's much better to use git to retrieve the files because it's >> independent to the host provider and we can use >> a lot of different transport protocols to retrieve the files (ssh, http, >> https, rsync, etc...). >> The problem when using git as a repository for binaries is that if you clone >> the repository you will download all the >> files (and all the history) which can be very very huge. Also replacing a >> package with a new one will not delete the >> previous files in the repository and it will always become bigger and bigger. >> This is why i wrote git-store that store only symlinks and not the binaries >> in commits. So cloning the repository is >> instantaneous. But the best thing is that the binaries are store in the git >> repository and the files can be retrieve >> (using the git command) on demand. > > cheers > > Erich > /Per ------------------------------------------------------------------------------ Dive into the World of Parallel Programming The Go Parallel Website, sponsored by Intel and developed in partnership with Slashdot Media, is your hub for all things parallel software development, from weekly thought leadership blogs to news, videos, case studies, tutorials and more. Take a look and join the conversation now. http://goparallel.sourceforge.net/ _______________________________________________ leaf-devel mailing list leaf-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/leaf-devel
