Mail Archives: geda-user/2014/03/11/12:42:43
John Eaton:
...
> A lot of designers do make copies of things all the time. A lot of
> commercial projects run into problems. Some designers have traced the root
> cause of these problems back to somebody using the wrong copy of something.
> That is why we want you to stop making copies, you are making our jobs
> harder.
>
> Everybody MUST keep their components in a repository. We call that "keeping
> a backup". It does not have to be on the internet and visible to the world.
> My personal backups are a monthly dump to dvd and stored it in the barn in
> case the house burns down. If you work for a company then you may work on
> a local copy of your files on your workstation but the "official" copy will
> be kept by the company in their repositories. You will make frequent (ie:
> weekly or even daily) check ins so that your latest work is always stored
> in a repository. That's how the pros do it.
I don't have that much hard experience to rely on, but I'm kind of fed
up on all copies of the same thing. I can also understand Алексей
Харьковский's urge to solve the things at hand, but I think we should go
"no more copies" route and start trying out different approaches to
integrate repositories in gschem sym file handling.
> Updating components do not break old designs. Once a component is released
> and used in production then you cannot make any material changes to that
> component.
That would require some best practicies rules and perhaps a program that
tests that thoose rules are not violated. It would be very nice if we
had some graphical regression tool. I know that lilypond.org are using
such things to verify that changes doesn't break the graphical output.
I can perhaps dig that up when time comes.
> You can however create a new revision of that component and
> release that one with a different version number. When a user does an
> update then they will see that there are new versions and can decide to
> stay with the old or replace it with the new.
That's another thing that should be handled.
...
> Besides versions there is a concept called variants. In IC design you want
> to create components that are flexible so the end user can customize them
> to their exact needs. We do this using parameters. But parameters do not
> work in two cases, they cannot modify a port list or a file list. These
> require variants.
>
> If you have a cpu with the option of having a debugger interface then you
> cannot select this with a parameter. The debugger will have io ports that
> don't exist on a non-debugged cpu. So you will release two variants, one
> with debugger ports and one without.
So to92 transistors pin djungle are different variants, and so are
different packages of the same mcu ?
For a chip, you can pick a subset of pins, e.g. power pins, and make
a separate symbol for that. That's not a variant, but what would that
be called, a "selection"?
Regards,
/Karl Hammar
-----------------------------------------------------------------------
Aspö Data
Lilla Aspö 148
S-742 94 Östhammar
Sweden
+46 173 140 57
- Raw text -