Andreas Pflug wrote:

> I created new makefiles for slony_funcs and xxid using the contrib 
> infrastructure, which makes them win32 compatible. IMHO the attached 
> versions should replace the current ones.
>
> After some short tests, Slony-q 1.0.5 seems to be able to replicate 
> with win32 servers using those modules correctly.

I hate to say this, but while it's nice to have the "proof of concept," 
I'd oppose putting this into the 1.0.x tree.  I'd much rather see it go 
into CVS HEAD / version 1.1.  Version 1.1 has a substantially different 
Configure/build system (which should make it MUCH easier to keep some 
separation from the PG sources!), and it's the "more experimental" 
place, so that Win32 support fits better there.

> The slon process currently refuses to link, the thread lib seems to be 
> the problem.

I'd be inclined to keep the patch out 'til this is addressed ;-).

> Apparently slonik uses fork(), which would make porting it much more 
> expensive. I won't spent more time on that, pgAdmin will do that anyway.

Somehow, I don't think Windows users would be too keen on the "little 
language" way of doing things of which slonik is a nice expression...

Taking that a step further...

The "good" way to use Slonik is to use other languages to 
generate/modify slonik scripts, whether that be via:

 1.  Using m4 to rewrite bits of a slonik script (I love the way that 
idea makes certain people feel ill ;-))

 2.  Embedding slonik scripts as 'templates' inside shell scripts

 3.  Writing Pearl scripts (or perhaps some more popular language 
starting with the letter "P"; yes, I intended that spelling ;-)) that 
generate slonik scripts as stdout output

Those are all pretty much anathema to the way things are done on 
Windows, with the result that it seems pretty reasonable for there to be 
Another Way on Windows. 

It would be nice to see a graphical "control tool", particularly if it 
also runs on Unix...