Time for a new DLL release

[Chris Faylor]

On Sat, Apr 29, 2000 at 07:04:19PM +0200, Michael Ring wrote:
>The main disadvantage is that a lot of overhead has to be installed in
>the first step before the gui for installation shows up. That is imho
>a mayor disadvantage. Another one is that the cygwin dll has to stay
>in memory and thus replacing it is a little more tricky (but not
>impossible) A static version of cygwin dll would help a lot if we
>would decide to go this way..

The first disadvantage is a whopper, IMO.  The GUI/RPM setup.exe would
be *huge*.

We have to deal with the second one in setup.exe, and so far I haven't
seen any indication that the method we're using is not working.

>Writing a 'pure' Windows installer makes it possible to create a
>single file with 'everythin included'. The problem of this approach is
>that if it's a 100% pure windows application that does not use
>cygwin's dll then there is a lot of code to be re-implemented to make
>the rpm databases available to the setup program. 8-(

Yup.  Catch 22.

>The best bet would perhaps be to create a windows installer that
>interacts with rpm via a shell-script. The installer could check for
>updates, rename/remove cygwin dll's if it needs to update the dll
>itself, download the files and then let rpm do the work. (I think that
>is pretty much simmilar to Mo's approach)

I've gotten a lot of interest in using Red Hat's GUI-based front-end to
RPM, though.  I guess we could shoot for that kind of "look and feel".

Another option might be a web based installer.

cgf