Hi Lester,

> You need to look at the source code for Win98. It is not a
> multi-tasking
> operating system. It can't even keep the time right under
> some conditions. If
> you want TCP/IP to work properly then all Win98 clients have to go.
>
> An effect I have just confirmed is the effect of running
> Local or TCP/IP access
> to a copy of Interbase running on Win98 ( it keeps a local
> copy of information
> for when BT can't provide Frame Relay ). Local access always
> runs fast, but is
> not thread safe, TCP/IP is about 5 times slower, but never fails.
>
> I method I use to keep the outstations alive ( all Win98
> machines because I need
> multi-screen ) is to run my own multicast messages, but I see
> the same periodic
> delays when Win98 is 'throwing a strop'. This is not a
> problem that changes to
> Interbase will have any effect on! That includes the message
> not getting through
> - which never happens to an NT client ( unproven, but
> reportedly the situation
> ).

I know the facts you're talking about. My message was mostly about problems
you may have with events notification on some platforms and under the
circumstances. But can you (or anybody else) give me a good explanation why
in my real-world example just two concrete events posted from exactly two
concrete stored procedures were always transferred slow and all others -
always fast? This is the reason why I was thinking about a possibility of
something working not very good in the engine/client in the case of Win9x
clients. Just assuming...

> Offcause we are going to change all Win98 to W2k - arn't we
> <g> - I haven't even
> got that stable enough to distribute yet, and it's obsolite!

Unfortunately, not developers but customers usually decide what OS should be
used for their purposes :-(
So there is still a need to support all of those damned Win9x.

Cheers,
Dmitry