5

Description of problem / environment

We currently have around 100 GB of CAD files (90k files, 6k directories) stored in a couple of Subversion repositories. It seems an unnecessary hassle / burden to keep this much binary data in Subversion. It's also a burden for people to check in new files as they need to add & checkout a directory before they can commit. The only "advantage", being able to just right click and "update", has the penalty of 2 copies of each file being stored (how svn works), and being very slow. There is no meaningful version history to the files - i.e., the CAD files are not modified further one they are added, or if so, in this particular case it is not data we care about - only the current, latest state, or HEAD...thus exporting the data out of SVN is straightforward. Editing the files is not really part of workflow and is more likely to be accidental, and it involves 5+ CAD systems so I'm not sure a "PLM" type system would really be ideal or warranted.

The current environment of the file server is Windows Server 2003 - that will likely change in 6 months time (either to server 2008 R2 + big RAID 6, or a NAS, probably server 2008 R2 involved either way)

Due to the sheer size, no one really checks out all the parts (or even a given directory) very often and there is already a read-only network share that updates itself once a day from Subversion. The auto-update process breaks all the time (svn working copy gets dirty or in a bad state and needs to be cleaned). That is how the majority of users access these parts so they are fairly used to accessing from share already, it's primarily a change to how parts get added.

What new workflow options are there? Am I missing anything?

I'd like to update our workflow for dealing with CAD files. The current on the table consideration is going to a straight windows network share. Ideally maintaining this read-only behavior, but obviously people need a place to dump new files and have them be added to the share. If the network share becomes the primary source of the data it will be important that people aren't opening, editing, and saving the files all the time. I suppose the importance of that is debatable, but generally if editing, the contract is they copy them to their PC so the "main" copy of a given file isn't modified for everyone else.

Is it not worth the hassle trying to separate adding files from accessing them? (to maintain read-only access of the share)

Setting the share to write but not modify isn't necessarily an option (if maintaining read-only is a core requirement), as CAD systems like Pro/ENGINEER Take CAD file XYZ.prt, and each save increment a number... Eg. XYZ.prt.1, XYZ.prt.2, etc, which will result in many copies if people are accidentally saving to the share.

So far I have a hazy idea that I can script something to handle a writable "drop box" that copies to the share, and for example...denies zip files, and refuses to overwrite any files. This leaves me with the manual duty of deleting any files (occasionally necessary, but rare - or could be given to a select group of users). Maybe despite its imperfection Subversion isn't terrible...I'm looking for some other opinions here. What I don't want is to change everyone's workflow only to make the situation more work for me or the users (30-40 users).

8
  • "There is no version history to the files" - huh? SVN isn't capable of doing anything except to keep the full version history. May 19, 2011 at 0:44
  • Obviously SVN keeps a history (I am well versed in that), but the CAD files are never modified (i.e. modification to a file checked in), only new files added. Any history that does exist in svn is not history we actually care about, only the current, newest state, or HEAD. This is not a typical use of SVN, which is one reason I'd prefer to move the CAD files out of it. I'll update my wording May 19, 2011 at 2:55
  • So to clarify, your users have a tool on their desktops (svn client) that they used for editing / uploading these files, and you separately have a read-only share where they can grab a (reasonably?) up to date copy of the current set of files. Is that an accurate description of where you are now? May 19, 2011 at 3:20
  • I'm not totally clear on the workflow you want: people check out a file, work on it locally w/ whatever CAD program they have, then save the modified file back? And at that point you don't care about the previous version? But until the new version "appears" back in the repository (SVN or new system), you want the previous version to be there for others to look at? Is that right? May 19, 2011 at 4:16
  • And what happens now if multiple people try to work on the same file at once? I assume SVN can't do anything to merge your CAD files... May 19, 2011 at 4:18

2 Answers 2

3

I should wait for your response to my comments, but...

How about:

  1. A read-only share for your files.

  2. A writable share with the same directory structure.

  3. When someone needs to update a file, they "check it out" from the read-only share - that is, they make a copy of it locally. (The limitation of what I'm about to say next is that there isn't a check-out process...)

  4. They work on the file locally, any temp files are only on their hard drive.

  5. When they're done updating the file, they copy it back to the correct directory on the writable share.

  6. Using one of the many file copy tools (rsync, SecondCopy, whatever), at whatever interval you want, files are copied from the writable directories to the corresponding read-only directory. The new version of the file will over-write the previous one, or you could keep versions at that point if you want to.

As I said, there's no actual check-out in this system, it doesn't deal with two or more people working on the same file at the same time. I guess the collision resolution could make use of the fact that people would have a local copy of their work (at least for a while) to fall back on.

1
  • Thanks - even those simple steps are helping me solidify in my head what needs to happen. May 19, 2011 at 15:13
2

I fail to understand why you are using a version control system when you are clearly not using it as such. Right now you are not getting the benefits of version control, yet are still paying for it in resources used.

Given your description I suggest using conventional file/folder shares instead. Why make it any more complicated than it needs to be? Structure is most easily established using clear and sensible folder naming and layering. I believe your users will not only adapt to it quickly but will thank you for the change.

1
  • It was setup this way many years ago, it was likely laziness because we are a software house and use Subversion for our source control (where it obviously matters a great deal!). There is zero reason to be using SVN for CAD files you aren't modifying, which is why I want to switch. I'm just kind of sanity checking that I'm making things simpler and not shooting myself in the foot Based on current read-only share behavior May 19, 2011 at 15:02

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .