On Tuesday 10 August 2004 5:22 pm, Brent P. Newhall wrote:
> Kristian Vandervliet wrote:
>> Brent P. Newhall wrote:
>>> Kristian Vandervliet wrote:
>>>> Removing "Save" as an implicit command has been floated by several
>>>> authors over the years. In theory it is a good idea; all data
>>>> becomes persistant. In practice it has practical problems; what
>>>> filename should the application use for the first save?
>>> "Untitled". Then "Untitled-1", etc.
>> That isn't very useful from a users perspective. File names should
>> convey useful information; "Untitled-4" doesn't tell me anything.
> It tells you that this is a document/project/whatever that you haven't
> explicitly titled yet. That's quite a bit of information.
Well, it's quite a piece of very specific information but it doesn't convey
any information about the contents of the file. The user would have to open
the file and scan the contents.
This is a minor point however. Solutions have been offered (E.g. open a
dialog when the application is closed and ask for a document name)
>>>> What if the user *didn't* want to save
>>>> their changes?
>>> Then they don't rename their document/project/whatever, and forget it.
>> Your solution only applies when the user is working on a brand new,
>> empty file. If they were editing an existing file, you've just
>> saved changes they want to undo.
> I'm confused. If the user is working in any document (new or existing)
> and makes a change they don't want, they Undo. The system still
> auto-saves constantly, but lets the user Undo as far back as they want
> (and is reasonable to store). I think I'm misunderstanding your comment,
> though; could you elaborate, please?
No, you understand it perfectly. The act of throwing your changes away is
probably rarer than saving your changes but it still happens. Would users
accept the extra effort of having to explicitely undo changes?
>>> I don't think that *perfect* undo is necessary, just good
>>> undo. There are a number of actions that are easily reversible
>>> without needing undo.
>> No, I think perfect undo is the only acceptable possible compromise.
>> I wouldn't want any OS to make changes I could not undo. It is
>> impossible to predict how a user may manage or edit their data; you
>> can guess, based on use-cases and scenarios, but even if you're
>> 99.999% correct you'll still be wrong 0.001% of the time, and with
>> enough users, that a lot of times.
> I'm confused again. Why do we need to predict how a user may manage or
> edit their data?
Well if we're going to talk about methods of managing data, I think we'd
better have some idea of how users will want to manage their data.
> I envision an Undo mechanism which provides complete Undo except in
> special cases for certain types of operations. For example, you can't
> Undo emptying the trash. In which case, the system provides the proper
> warnings, and if the user still didn't really want to do this, then, well,
> the user screwed up. I don't see why these sorts of cases completely
> negate the advantages of automatic saves.
>> Users data is far to fragile and important to run the risk of damaging it.
> Then why do we write applications that constantly run the risk of *losing*
> that data?
The problem for me is that we have two imperfect systems; one where the user
must implicitly save their changes, but where they risk loosing their data,
and one where the system will save their changes but they risk damaging their
data. In the first model, the user screws up. In the second model "The
computer screwed my work up again! Aaargh!".
Having to go back, find the last good revision and roll-back is going to be
time consuming for the user. Would users be willing to do it?
> I would rather have a aystem that constantly saves my data with a small
> chance that I may have to manually Undo some accidental change on my part.
I would probably prefer a system where I must implicitly save my changes but
which *also* has the ability to roll back to previous versions. Versioned
files are nothing new; VMS has being done them for decades. Believe it or
not ISO9660 supports versions. They're a good idea, especially in todays
world of cheap and abundant storage. I'm just not sure I'd want my computer
to try and save me from myself by saving all my work for me :)