John Levon wrote:
On Thu, Sep 14, 2006 at 04:14:46PM +0200, Helge Hafting wrote:

Well, it can be automatic of course.  Still, there are times when
it is useful *not* to save.  The cat walked across my keyboard,
but it doesn't matter for I wasn't going to save this.
Well, we could remove "save" and keep "revert", or even rely on
"undo" for this sort of thing.

/Now/ you're starting to think seriously :)

A decent undo system is indeed the way to deal with these issues. The
book expands on this idea at length.

Yep. Unfortunately for really decent support we need some facilities
within the desktop environment...
We really shouldn't depend too much on a "desktop system", there
being many to choose from.  Different os'es, different
window managers and so on. Independence is a good thing here,
"do one thing and do it well".  LyX process documents really well.

For "desktop software" my preference is something that is small,
get out of the way, and simply let me use the other software that
do real work.  A handful of xterms really is my idea of a nice
user interface for most purposes - the command line is *fast*,
no waiting on a pretty GUI.  There being almost no "desktop" to
"integrate" with.  Of course many others have different preferences,
but we don't all have to use similiar desktops.
Not really, they just need to know that the stuff in the window
isn't persistent.

Take a step back and think about how absurd that really sounds.
Absurdity depends on where you come from.  Someone who used to
do office work with real paper, might like the computer desktop to
be very much like a real desktop. If you write something somewhere,
it stays unless you erase it or throw it in the bin.  I understand that
someone with such a mindset find it absurd that a paragraph
of text can disappear if they don't take active steps to preserve it.

But I have a different mindset.  Word processors have been there
almost all my adult life. Well, I wrote some stuff by hand in my student
days - portable computers were too expensive.  I rarely use paper now.
Email, web pages, and pdf files is what I use.  I know how computers work.
And more important - that is what I am used to. Not paper. Now, having to save
the document is an inconveniece because one have to press two keys,
but there is nothing absurd or unexpected about it for me.
And if you think of it - most screen text really is transient.  All
that typing on the command line - it doesn't get stored anywhere.
Only the effects of the commands remain.

A lyx without "save" would be convenient: one less thing to do, more
free space in the menus. The question is, is it worth the development
effort? We might have to add a "editing log" to the file format.  That
is theoretically easy - lyx already get its actions from the user interface, we
would just have to save them. And replaying them is also easy - lyx already
have all the edit functions, just get the input from the log instead of
mouse/keyboard. Still, it has to be written, and then it has to be supported
over format changes as lyx evolves. That last part could be "interesting".

Seen isolated, no save is needed here.  Lyx can "save" on close,
as well as after 5s of inaction to be on the safe side.

But I have thought some more about this, and I see problems.
If I send the document somewhere (mail, ftp, scp, copy it
to another folder) then I want to know what I get.  Most other
software will indeed grab a "file" and send it off somehow.

This is the sort of thing I mean by DE support.
And how would that support work?  Having each program
know how all the others work just in case they ever
need to cooperate is clearly out of the question.

It is also currently hard to know which documents might
be open and by what process.  Linux has "lsof", but
the document might not be on local storage.  I have two
computers here, and might edit on one and run
the mail software on another, accessing a common file server.
Unusual setup perhaps, but a user who don't grasp
"file systems" wouldn't see how this might cause its own
special issues either.


The email program may be up and running, the message written
already.  I put some finishing touches on the document and may
be able to attach the file and send it faster than the above mentioned 5s
autosave.  After all, I am good at sending attachments.  So, did I send the
document with the latest changes or not?

But your email client is plugged in - it knows you've asked for "the
current version of the file", not some previous snapshot. It'll show you
a preview of what you're sending.
I simply do not want to page through a 12-page preview of the
document I'm going to send.  And the email client is not supposed
to know all those formats.  Sure - it could use os support to
preview unknown files.  But I don't want to look.  I know
that the document was ok, that all I needed was to delete
the stupid offending joke at the end before mailing it off. ;-)
Besides, many kinds of files don't have a sane "preview".  How do we
preview a database file, for example?  I assume you want the same
interface with "no save" there too.  Update some records, mail
it off, then do some radical changes.  The mail software had better
get this right.
Lyx can change, getting rid of another menu entry.  It still have to
interoperate well with other software.

In all seriousness, I don't think we can implement something like this
unilaterally. To my lasting sadness.
Well, the idea is interesting, although it looks like much work to
loose just one menu entry.

Helge Hafting

Reply via email to