topbanner_forum
  *

avatar image

Welcome, Guest. Please login or register.
Did you miss your activation email?

Login with username, password and session length
  • Sunday December 15, 2024, 3:16 pm
  • Proudly celebrating 15+ years online.
  • Donate now to become a lifetime supporting member of the site and get a non-expiring license key for all of our programs.
  • donate

Author Topic: Which to use when? Reparse points (Junctions), Symbolic links, and Libraries.  (Read 3244 times)

IainB

  • Supporting Member
  • Joined in 2008
  • **
  • Posts: 7,544
  • @Slartibartfarst
    • View Profile
    • Read more about this member.
    • Donate to Member
Since getting xplorer² years ago, I have tended to favour using reparse points (folder Junctions) quite a lot to organise my data into more readily-accessible libraries. I don't really use symbolic links though, as I tended to follow the xplorer² author's (Zabkat's) advice. I am also learning to make better use of Windows Libraries too, since migrating to Win7-64 Home Premium.

But Zabkat seems to have been reviewing the situation and has changed his opinion somewhat about reparse points and symbolic links. There's an interesting couple of posts at the xplorer² blog:

Shades

  • Member
  • Joined in 2006
  • **
  • Posts: 2,939
    • View Profile
    • Donate to Member
Fun thing with symbolic linked stuff...I use them under Linux (Ubuntu 10.04 Server edition) and my file-manager of choice 'Midnight commander', but it is always tricky when you want to remove them. it is very easy to not only remove the link, but the file/folder as well.

However, I do like to use them for making stuff available (within limits) in a 'walled-off' folder structure. I don't lose any sleep over the access to or the loss of content on that PC at all at any given time.

NTFS has proven to be more problematic to me. Specifically the NTFS version in Windows 2008 R2. When accessing it a lot through network for an hour or so (SOAP calls to Oracle database server), NTFS just told me that it was not able to access the hard disk anymore. And with a lot, I really mean a lot, as I was trying to mimic the traffic a computer with 64 Itanium processors needs to process (on a quad-core i7).

I lost two 350 GByte databases that day, as Oracle couldn't make heads or tales anymore from the mess NTFS had created. Even the archive logs from Oracle were damaged. And the funny thing was that I had an older version of one of these databases running on a (AMD single core) 2003 Server and that one was able to complete my test without any problem for any amount of time while not really being slower at all.

Because of this incident I do not have the same confidence in the latest version(s) of NTFS as in the older NTFS version(s)  from 2003/XP.

Back to the topic again. Directory opus ( and higher has functionality comparable with the Libraries in Windows and I already dislike it in DOpus, so don't expect me to use it in Windows 7/2008. Never worked with folder junctions, so no (good/bad) comment from me on that matter.

An intelligent (to me at least :)) folder structure with symbolic links always made more sense to me. More fine-grained control, (likely) more speed gaining access to the accumulated data and (likely) faster updates from file changes. But that could all be just me.

f0dder

  • Charter Honorary Member
  • Joined in 2005
  • ***
  • Posts: 9,153
  • [Well, THAT escalated quickly!]
    • View Profile
    • f0dder's place
    • Read more about this member.
    • Donate to Member
Because of this incident I do not have the same confidence in the latest version(s) of NTFS as in the older NTFS version(s)  from 2003/XP.
I would blame Oracle before I blame NTFS - seeing the amount of data-loss bugs and just plain weird behavior I see discussed on our companys internal mailing list, I'm surprised anybody feels safe using that in production.

NTFS is put under extreme stress testing by Microsoft. They're running 256-core machines with several terabytes of RAM, for chrissakes :) (there's a channel9 video somewhere where Mark Russinovich, iirc, logs onto that beast). And John Carmack has played with content-addressable data storage on NTFS (if you look a bit into the amounts of data required for content in their latest engines, well...)

Anyway, not saying it's impossible that there could be bugs in NTFS - just that given the choice of doubting NTFS or doubting whOracle, it's a no-brainer where I'd put my money :-)
- carpe noctem