I’m no great lover of smoothwall as I just began using it, but I found it to install pretty cleanly on a fairly recent socket 775 running a PentiumD processor. Not the newest machine for sure but not the oldest in the bunch either.

TimT

It is not true that the whole team resigned. In 1996, two major product development projects started, that used Erlang as its core technology. This is also when OTP was born, and the responsibility for Erlang was transfered from the Comp Sci lab to a more regular Ericsson design unit. Of the people who left for Bluetail, most were CSLab, and a few were at the OTP unit. There were enough Erlang experts available to replace those who resigned (who were actually involved in the main Erlang/OTP development), and the show could go on.

Also, AFAIR, Nortel didn’t fire most of the Bluetail team. Some, who were not actively working on the product, were let go. By now, I think they’ve all gone on to other things, but a significant part of the Bluetail gang developed several products for Alteon/Nortel first. Kreditor, Tail-f and Synapse were three companies started by (or with great help from) ex-Bluetail people.

BTW, CSLab, after transfering Erlang into production, went on to research other things that unfortunately didn’t see the light of day. They did some tremendously exciting stuff code-named “Communication machine”. Exaggerating a little bit, it was the IPhone 10 years ahead of time, with clunkier hardware.

Indeed there is something wrong with C++, just as there is something wrong with every language. It is the wrong tool for many of the types of project that large telecommunication companies run. I think using C++ for building complex distributed soft-real-time products in projects with several hundred developers is simply asking for trouble. I will stipulate that small teams of very experienced programmers can do almost anything in C++, but what this too often translates to is that they build a brilliant prototype that is then transferred into a large project with mere mortals who will find a hundred different ways of shooting themselves not just in the foot, but in all parts of the body.

One tiny observation: Since you don’t expect to buffer unknown messages, it will be slightly more efficient and semantically clearer if you write

case next_message() of
{done, PrimeNumbers} ->
…
end

next_message() ->
receive
Msg -> Msg
end.

This way, the process will crash if you get an unexpected message, rather than buffering it and slowing down subsequent receive operations.

There are of course many ways to overwhelm a receiver even in Erlang, but of course you know that. A subtle point is that the reduction-counting scheduler will actually punish the sender with extra reductions if the receiver’s message queue is very large. This slows the sender down even more. While this was a ‘clever’ optimization on single-core, it is not obviously a good idea on multi-core, since it implies synchronous delivery and locking.

Making any assumptions based on reduction count scheduling was always strongly discouraged. In an SMP system, such assumptions are /guaranteed/ to break down.

Thanks for detailed explanation.

I am trying to do a vice versa of example, i.e trying to convert details from XML file to csv file.
I am not sure how to set the mapper for XML file and read them for flat file.

Could you please provide some details or possible share the code.

Thanks

Is the binary data organized in - for example - (fixed sized) records? Are the bytes you need to replace, interpretable in the context of some high-level structure? If that is the case, then it might be an idea to read binary data, convert it into higher-level objects (some form of records), update the objects and then write them out to another binary format or the same format.

On the other hand, if nothing of this makes sense it’s better to write a dedicated program.

You might of course still use Spring (Core), assembling the various modules together BinaryFileReader / BinaryTransformer / BinaryFileWriter.

That said, I do accept your criticism. Your posting matches my own thoughts in that the problems made me look unprepared and really distracted from the main messages that I was trying to present, tuning from first principles.

Regards,
Kirk

Your kvm-manage scripts seem to be really cool. Good work.

I haven’t heard about VDE2 before, it seems to be very useful for cloud like virtualization deployments. Thanks for the tip, I will investigate.

Merry Christmas,
/jens

this looks like a very nice step by step howto. I also like kvm and I use it for several things. I am coming from Germany and I have recognized one problem with the -k and -vnc option. I have selected a German keyboard layout, but when using the vnc connection, some keys are missing, i.e. the pipe, the keypad, etc.

I saw that you use sv as layout. Did you recognize such problems, too?

And one suggestion: Have a look for vde2. Cool stuff.

I have written a kvm-manage script that can start and stop (and some more) kvm guests controlled. That means. One-by-one with time-parameters. It is desigend for the init.d. If you want to, have a look here:

German blog entry
http://www.roessner-net.com/?p=219

and the script (commented in english):
http://www.roessner-net.com/kvm-manage

So long
Christian