Random musings... (Was Re: Long filenames in DOS server)

Hi,

Gavin Thomas Nicol wrote:
>
> This is a good question. Perhaps we should have a CVS repository somewhere?
> How about hosting VSTa development on sourceforge?

Sourceforge would be a pretty good bet, although an alternative might be
just to run the sourceforge software on the current server? From what I
can see, this would be pretty straightforward too (given that a server
already exists).

Personally I've already gone for sourceforge for my Liquorice project
(it saves me a lot of irritation not having to do lots of admin work).
The only downside I've seen is that late in the evening UK time (2300ish
say) sourceforge sometimes goes a bit slow (must be the Aussies waking
up - as if winning test matches wasn't enough now they're running away
with all of the bandwidth ;-) well perhaps not...)

FWIW I'm looking at doing some VSTa network hacking very soon as my code
is headed in a lot of similar directions, and having I'll be interested
to see how it's new TCP/IP stack works with VSTa (I now understand why
Andy ported KA9Q and didn't start from scratch - although it is quite an
interesting thing to do :-)).

I'm going to have to have a good look at the server proxy code as I
think this offers a lot of really useful possibilites to the code I'm
doing (which is currently targetted at big 8 bit micros - 32 kBytes of
RAM is lot!). Ideally I could use VSTa to perform a number of services
for the AVR microcontrollers - effectively the same thing MS are after
with Unversal plug and play but *much* simpler to implement and without
MS-levels of overhead.

The environments with which I'm interested typically have about 100 or
so microcontrollers and I'm working on replacing the RS485 networking
with Ethernet/IP/TCP. Has anyone any idea how well the KA9Q-derived
network service will handle 100 connections? I've never tried this many
before, but I'd hope it shouldn't be too much of a problem.

[Aside: What I'm looking to do is use VSTa as a host for streamed data
(each microcontroller sending at a slowish rate of say 1 kByte per
minute) with each stream being compressed and stored in a parameter
database (about 50:1 compression is possible - I've already been doing
this for 2 years). As I'll now have a peer-to-peer network rather than
RS485 I'd plan to provide another VSTa service that allows the
microcontrollers to identify others with interesting characteristics (a
registry of sorts). There are other browserish things that need doing
too but I think they can wait for a while!]

How easy would it be to extend the network server to implement a
different transport layer? I like the idea of the IL transport that
Plan 9 uses (although I must admit I've not looked at the code for it)
as it sounds much cleaner as far as implementations on small targets go
- TCP is a real nuissance. I also seem to remember Andy thinking about
implementing a raw Ethernet protocol for remote-message-passing (as with
QNX 4) - any more thoughts on this?

                        Regards,
                        Dave
Received on Mon Jul 17 13:59:46 2000