Ok, I was mainly talking about forward compatibility but even writing a previous version of a file format out requires:
1) you know automatically what is changing between versions 2) you know how to identify and gracefully omit objects, operators, connections and data that will not be recognized in the older format As stated before the persistence code in XSI is not centralized and therefore it is still a significant effort IMO to do 1) and 2) above so I guess we differ on the definition of "much simpler". ;-) There are also cases where we rewire/reorder graph connections or version/replace operators to fix issues so we would also potentially need to rewire things or write out different/older versions of operators when saving the old format. In my opinion the save-previous-version strategy only works well for standard file formats which are precisely defined and change infrequently. (internal file formats do not usually meet this requirement) The simplest thing I can think of is to have a way to override the safe behaviour and let the user import the file at their own risk. (equivalent to changing the version number manually in a Maya ascii file) -- Brent From: [email protected] [mailto:[email protected]] On Behalf Of Luc-Eric Rousseau Sent: 18 April 2012 01:15 To: [email protected] Subject: Re: 2013 save scene = no load in 2012? What we said was that allowing the software to load a scene from a future version is very difficult. Adding feature that allows the software to save a file compatible with one version back is a much simpler problem. Much less to test, and no time machine required. On Apr 17, 2012 5:01 PM, "Eric Turman" <[email protected]<mailto:[email protected]>> wrote: But I was originally suggesting the prior...the one that Luc-Eric said that "the team could conceivably consider." On Tue, Apr 17, 2012 at 6:56 PM, Steven Caron <[email protected]<mailto:[email protected]>> wrote: nope, mixed up again... 2013 could in theory save a file knowing which features 2012 doesn't support, but 2012 would need to radically change how it loaded data from future (forward) versions and features it does not yet know about. brent was talking about the latter. On Tue, Apr 17, 2012 at 4:52 PM, Eric Turman <[email protected]<mailto:[email protected]>> wrote: Wait...I thought that Brent said that it was not trivial to implement a 'save as previous version'; It sounded like it would take considerable resources to implement that. -- -=T=-
<<attachment: winmail.dat>>
