5 ms·
>https://arstechnica.com/gadgets/2017/11/microsoft-and-github-team-up-to-take-git-virtual-file-system-to-macos-linux/ https://arstechnica.com/gadgets/2017/11/mi
by trhway 8y ago
>https://arstechnica.com/gadgets/2017/11/microsoft-and-github-team-up-to-take-git-virtual-file-system-to-macos-linux/ https://arstechnica.com/gadgets/2017/11/microsoft-and-github...
sounds like MS is trying to make Git into Clearcase. Typical bunch of FUD stemming from not understanding the tool and how to work with it, and of course the monstrosity of a solution ( "Git at enterprise scale" (https://www.gvfs.io/) https://www.gvfs.io/)) to the rescue:
"Microsoft wanted to move to Git because of Git's features, like its easy branching and its popularity among developers. But the transition faced three problems. Git wasn't designed for such vast numbers of developers—more than 3,000 actively working on the codebase. Also, Git wasn't designed for a codebase that was so large, either in terms of the number of files and version history for each file, or in terms of sheer size, coming in at more than 300GB. When using standard Git, working with the source repository was unacceptably slow. Common operations (such as checking which files have been modified) would take multiple minutes.
The company's solution was to develop Git Virtual File System (GVFS). "
I think it is more that Team Foundation Server is failing to scale and MS needs [their own branded/controlled] "enterprise-scale" VC solution for themselves as
well as for their customers. GitHub with GVFS on top (ie. kind of Sharepoint for VC) fits nicely here.
- twunde 8y agoI think the more likely benefit is that pretty much every engineer knows git, it has become the de-facto standard even if other alternatives like Mercurial are easier to learn and use. In terms of onboarding new engineers, this probably saves them a week where new engineers don't have to ramp up on a new technology. It might make a difference in terms of recruiting new engineers. In terms of understanding their customers, I'd guess that a significant number of customers are using git. It probably also makes it much easier to open-source code.
- WorldMaker 8y agoFrom what I've heard, TFVC would have scaled just fine, and was similar enough to the internal to Microsoft ancient Perforce fork called Source Depot, but it wouldn't have been particularly agile. TFVC like CSV/SVN has a harder time branching because it is branch-as-path/copy-on-write, for one instance. Scaling a repo that size requires intricate lock systems which TFVC supports, but aren't very developer friendly, etc, as the Windows team was already fighting with Source Depot. The issue isn't that TFVC wouldn't have scaled, it's that Windows developers wanted git for many of the reasons that most of us use git: cheap branching, distributed/lock-free code development, etc. GVFS is nothing like Clearcase/Perforce/Source Depot/TFVC. It's still git, it's just offloading some of the .git object database on a machine to servers, just like you might not download every file in your Dropbox/OneDrive account to a local machine. GVFS doesn't even have to be "centralized". GVFS uses a simple REST API to query object servers and I still don't think anyone has had the need to do it yet, but building a GVFS daemon to pull objects from IPFS (and pin objects to it) is entirely possible.
- trhway 8y ago>TFVC would have scaled just fine is pretty contradicting to >TFVC like CSV/SVN has a harder time branching because it is branch-as-path/copy-on-write, for one instance. Scaling a repo that size requires intricate lock systems which TFVC supports heavy branching and locks is exactly the things which make system less scalable, in all dimensions like devs number, repo size, parallel product versions being developed etc. >Windows developers wanted git for many of the reasons that most of us use git: cheap branching, distributed/lock-free code development, etc. exactly my point - they hit scalability limits of TFVC. You just said it using different words. We all using Git because it is [architecturally] the most scalable VC out there (i did development of a major [back then] VC almost 2 decades ago - we were starting to develop bunch of ideas back then which ultimately came out to the wide public in git)
- WorldMaker 8y ago> heavy branching and locks is exactly the things which make system less scalable, in all dimensions like devs number, repo size, parallel product versions being developed etc. It's not a contradiction. Those things scaled just fine for decades of software development. They aren't "nice", aren't "fun", but they are a scaling solution. The Windows team used Source Depot ("Perforce") for something like 30+ years, at the scale of the Windows team. There are companies and repositories in the wild in Fortune Xty companies with equally old, equally "scaled" repositories in classic, ugly centralized lock-based repositories. You and I may not want to work on any of those repositories, but they absolutely scaled in historic fact for a scale that still dwarfs the lifespan of git. You may confusing workflow flexibility with scalability? Command-and-control workflows are very scalable, but not very flexible. Distributed workflows are very scalable, and very flexible, when built right. The Windows team wanted a break from classic Waterfall and Command-and-Control workflows. They knew how to scale those and had scaled them successfully for many years. They used the change in source control as an opportunity to change workflows. TFVC could have done the job if they still wanted that workflow (presumably better than the hodgepodge of very old Perforce code and Perl scripts that was Source Depot, that had "scaled" for decades). There are still companies and developers happy with that workflow. The interesting part is that Microsoft wanted "something like git", that enabled those new workflows, but that scaled to what they were already doing in centralized repositories with Windows. Instead of investing in a Distributed-TFS, they decided to just use git, and they put a lot of effort into performance tuning to git which was contributed upstream and which we are all benefiting from now. Yes, they also built a simple optional plugin (GVFS) that is itself open source, and built on simple open REST standards. It's pretty much the exact opposite of building "their own Clearcase" or whatever; they didn't build a D-TFVC for their needs, they contributed directly to the open source one in an open way, exactly as you would want them to.