Please create an account to participate in the Slashdot moderation system

 



Forgot your password?
typodupeerror
×
Technology (Apple)

Journal XMLsucks's Journal: Apple's innovative Time Machine 4

On Monday Apple announced an upcoming Mac OS X feature called Time Machine. This is a very innovative feature, and one which the world has desperately needed. Several backup products already exist (I use pdumpfs for its sparing use of disk space via hard links), and thus you may wonder why I find Time Machine so innovative. Not a single person has pointed out the innovation, and Slashdot's commentators completely missed out, and worse, many ignorantly criticized Apple for copying VMS's versioning file system when Apple's solution is nothing of the sort.

First, Apple's solution isn't a versioning file system. A versioning file system is good for revisions, but not for backups, because a versioning file system normally stores the revisions within the same file system (to shared disk blocks) and thus on the same disk. For backups, you want the files on a separate disk, to survive disk failures. Apple claims that they put their files on a separate disk, so that rules out a versioning file system.

For the backup, Apple implements a traditional periodic, automated backup system. It copies the files to an external disk. The small fraction of people identified by Apple that already backup their files probably use this approach.

Apple's innovation, and it is massive innovation, is related to the restoration of data. Not the restoration of files, but data. The problem with backups is that they operate on files, rather than the units of data that interest people. For example, an address book may store all of the database records within a single file; if you want to restore a single record from a backup, you either must replace your current file with the older file, thus losing all of the intervening changes, or try very hard to extract a subset of the records from the old file, and merge them into the new file. The file restore experience is very frustrating. Apple provides the ability to restore subparts of a file, to merge into your current file. This is fantastic. You can see it in action on their animated demo: they restore a single address-book record, and they restore a set of photos to iPhoto. In both cases, the applications work with Time Machine to recover old data, and to merge that old data with the current version of the file. And it does more: in the case of iPhoto, it maintains consistency across all of the files that store the data; iPhoto stores images as individual files, but stores their meta data within a central database, and thus the restore process recovers the old pictures, and adds their meta data back into the database. None of this is possible on the operating system level because it only understands files --- Time Machine requires support from the applications. Apple announced that they provide developer APIs to integrate the Time Machine into all applications. I can imagine OmniGraffle permitting restoration of a layer, or a particular graphic, from an old version of a file into the current version. Or for Microsoft Excel to permit restoration of a single sheet or graphic of the spreadsheet. For applications that ignore Time Machine's APIs, you'll be stuck with boring whole-file restores.

This is even better than normal undo-redo histories of applications. How many applications let you selectively undo or redo an action from an arbitrary location in the stack? This Time Machine lets you restore a record of data from an arbitrary location in the stack (within the extent of data integrity).

This discussion has been archived. No new comments can be posted.

Apple's innovative Time Machine

Comments Filter:
  • That sounds nice, do I need a terrabyte to record different versions of my files and apps from a 160 gig base system?
  • Got here via the link in your MacWorld forum post ...

    Apple's innovation, and it is massive innovation, is related to the restoration of data. Not the restoration of files, but data.

    That's an interesting observation, overlooked in nearly all the relatively mundane commentary/discussions of Time Machine I've seen so far.

    The problem with backups is that they operate on files, rather than the units of data that interest people.

    Obviously there are exceptions, but in general I think computing interfaces

  • Excellent post. I think nearly everyone missed it; I know I did. I suppose that's what happens when you make everything appear seamless in your presentation.

    As are many other people, I'm concerned about the storage requirements. Some data could be compressed I suppose, but a lot of data which concern people the most (e.g., audio, image, video) don't compress a lot. At present I have a single 500G external that I use to do (in effect) image backups of my 500G internal. (All right. I admit it. I use Sup

    • by klaiber ( 117439 )
      I'm actually not too concerned about the space requirements. I use Dantz Retrospect to backup all the machines in my household to a Mac Mini "server". The backups are incremental and run every day (just like Time Machine, I exclude ephemeral files like browser caches). Every few months I rotate among two 400GB drives (one of them is always offsite). While I don't do video editing, I do shoot a lot of RAW images (5MB a pop; editing in Photoshop can blow that up to 50MB or more each), and I have yet to ov

"Engineering without management is art." -- Jeff Johnson

Working...