Yeah, my point isn't that replaying the commands is currently a viable
alternative for scene recovery (though I would say it's better than
nothing in the cases where the auto-recovery fails).  Just that it
could be a really good solution, if some more effort were put into
ensuring every interaction got logged as a command of some kind.  And
since a lot of people rely on command logging as a way to build
scripts, they'd be killing two birds with one stone.  The plugin is
just a proof of concept.

A journaling approach just seems like a better way to deal with file
recovery than a weird last-second binary dump, especially given that
the dump often contains the thing causing the crash in the first place
and has no way to handle things like power outages, auto-logouts, etc.

There's no reason 3D software can't be as robust with file recovery as
Nuke is.  Even if there are commands that can't be logged for some
reason, you could have a special dirty state that triggers a normal
auto-save in those instances to ensure that the journal-based recovery
can still work properly.

On Fri, Sep 6, 2013 at 8:58 PM, Matt Lind <ml...@carbinestudios.com> wrote:
> My only beef with your plugin is it cannot account for commands which do not
> log.  There’s a good probability the scene your plugin generates is not an
> accurate representation of what last state of the scene actually was before
> it crashed.  Animation edits, for example, do not log at all.  Custom tools
> flagged to not log, or tools called from self installing commands do not log
> either.
>
>
>
> Back in good ol’ days of XSI v6.x when we were treading water to get
> anything to function in XSI without exploding, I desperately tried to
> salvage crashed scenes using a similar technique, but because many commands
> were not logged it was not possible to salvage work or even rebuild it
> enough to send to Softimage to diagnose the cause of the crash to get it
> fixed.  Critical missing steps caused the rebuild script to error out, or if
> it was lucky enough to get to the end without error, the end result was not
> at all like what it should’ve been.
>
>
>
>
>
> Matt
>
>
>
>
>
>
>
> From: softimage-boun...@listproc.autodesk.com
> [mailto:softimage-boun...@listproc.autodesk.com] On Behalf Of Andy Jones
> Sent: Friday, September 06, 2013 8:44 PM
> To: softimage@listproc.autodesk.com
> Subject: Re: Would you like to recover your scene?
>
>
>
> I've said this before, but the script log comes really really close to
> implementing "journalling" which is the thing that made filesystems more
> robust in the 2000's.  the binary dump emergency save thing is really kind
> of a silly way to attack the problem of replaying a journal of operations.
> If Autodesk realized this, they'd prioritize the completeness of command
> logging and build a simple toolset for replaying the unsaved operation
> journal.
>
>
>
> I sent my "repeatHistory" plugin to the beta list a while ago and got
> crickets.  Maybe I'll try again...
>
>
>
> All it does is parse the script log for the last open or save operation and
> them exec the remainder.  It would work even better with an event that sets
> the script log path on scene open/save.
>
>
> On Friday, September 6, 2013, Jeremie Passerin wrote:
>
> Got issue with the auto-recover lately, but the scene was actualy properly
> saved before crashing. just needed to load it manually.
>
> http://xsisupport.com/2011/10/15/crash-recovery-in-softimage/
>
>
>
> On 6 September 2013 13:18, Eric Thivierge <ethivie...@hybride.com> wrote:
>
> You're doing it wrong...
>
>
>
> On September-06-13 4:15:36 PM, Eric Lampi wrote:
>
> SoftImage: "Hey Eric, I see that you crashed.. How would you like to
> recover your scene?"
>
> Eric: "Sure that would be great! Go right ahead, bring it on back!"
>
> SoftImage "You'll get nothing and like it!"
>
> Meh
>
> Eric
>
> Freelance 3D and VFX animator
>
> http://vimeopro.com/user7979713/3d-work
>
>
>
>

Reply via email to