Re: A riskier alternative to "latest"?

[Chris Faylor <cgf at cygnus dot com>]

On Fri, May 12, 2000 at 05:26:02PM -0400, DJ Delorie wrote:
>
>> This is similar to a lot of other projects so I don't think this is a
>> very radical concept.  The only thing I don't know about is what to name
>> the directory, actually.  Is "development" clear?  Some projects call it
>> "dontuse" or "new", too.
>
>DJGPP uses "alpha" and "beta" subdirectories for stuff like that; we
>don't get too many complaints (I don't remember *any*), but the
>install tools don't scan those directories anyway.  They key is to not
>install the test versions *by default*.
>
>If setup doesn't go more than one directory deep, we could add
>alpha/beta directories within each package.  Or, we could add
>alpha/beta siblings to latest; setup should ignore those also.

>I think the concept of alpha/beta is pretty well understood.  I don't
>see why we'd need to invent some other term.  Alpha is for things that
>probably won't work, beta is for things that probably will work,
>latest is for things that do work.

The reason I didn't suggest "alpha" and "beta" is that the tools in
latest are already supposed to be "beta" and I thought that this might
confuse things since we would have a beta version of our beta version.

>Another option is to allow tagging individual tarballs with "risk
>factors".  To do this we'd need either a rock-solid versioning/naming
>scheme, or start using some master config file for setup to read, so
>that setup could prompt for "do you want to try experimental
>versions?"  and do the right thing.  Of course, we'd need a way to
>revert to stable versions if they break.

It would be nice to have setup ask if the user wanted to try experimental
versions.  I don't envision that things like "ash" will be experimental
for long, though.

Maybe the best plan is to just release ash and a vfork version of cygwin
and drop back if this causes problems.

cgf