[DotGNU]Re: dotGNU question

[Norbert Bollow]

Jon,
  the DotGNU bytecode might end up being very similar to Java
bytecode, so that only minimal modification would be necessary
to some existing Free Software for generating and executing this
bytecode.

However there is a reason why a new bytecode language (or at
least a modification of an existing bytecode language) is needed:
At least in the vision for DotGNU that I have proposed, there is
not just a "virtual machine" that runs on a single computer, but
two layers:  a "DotGNU secure execution environment" (which is a
rough equivalent of the "virtual machine"), and another layer
called "DotGNU Distributed Execution Environment", which will be 
built on top of the "DotGNU Secure Execution Environment" layer.
The "DotGNU Distributed Execution Environment" will really be a
rebust operating system for "distributed virtual internet
servers", the kind of set-up that you want if you have multiple
internet servers (perhaps in different parts of the world)
that all need to have read/write access to a common database,
and that all need to remain able to make database transactions
even when there are temporary problems with the network
connection between these servers.  The DotGNU Distributed
Execution Environment will come with a lot of features to
support the necessary technicalities such as e.g. database
transaction journalling and roll-back.  In order to implement
these technicalities, some parts of the functionality will
reasonably by implemented taking into consideration the
peculiarities of the underlying OS.  This code would be made
available for execution from the bytecode via something like a
"system call".

The DotGNU Distributed Execution Environment architecture will
eventually be used for example by the DotGNU Virtual Identities.
Microsoft's centralized architecture for the passport system is
really a poor alternative to this approach.  A truly distributed
system (like DotGNU is planning) can be much more reliable and
robust, but in order to make it work well even under heavy loads
in the real world (where network problems happen) it must be
built on top of a rebust system like the DotGNU Distributed
Execution Environment.

HOWEVER work on implementing DotGNU Virtual Identities is
starting in parallel to the work on designing the DotGNU Core
Platform, DotGNU Secure Execution Environment and DotGNU
Distributed Execution Environment.  There will be a
quick-and-dirty first implementation of the DotGNU Virtual
Identities project (using tools that are available today) so
that we will be able to start competing with Microsoft for
market shares soon.  That will send a strong message to the
world that we're serious about this project, and a force to be
reckoned with.

Unless Microsoft screws up royally (it is possible that that may
happen, because with the architecture of their "passport" system
they have built a single point of failure into their system), it
is really impossible to win the market shares war against
Microsoft by merely doing essentially the same things that
they're also doing.  Mono may turn out to be a smart business
move for Ximian, but Mono will not stop Microsoft's Hailstorm.
If Free Software developers create a Hailstorm clone that is ten
times as reliable as Microsoft's product, that is not going to
stop the success of Microsoft's Hailstorm.  It is a key element
in Microsoft's strategy to be perceived as being the innovator,
and Free Software merely imitating them.  As long as Microsoft
succeeds in upholding this perception, they can win the market
shares war and end up with an effective monopoly on e-commerce
authentication services -- a monopoly that they might be able to
leverage to gain control of other key markets, such as the
market for e-commerce server software.

This is why it is not enough to simply clone (parts of)
Microsoft's .NET initiative and provide a different
authentication service.  We must become the innovators and
create a framework (the DotGNU Distributed Execution
Environment) that major ecommerce sites will want to use because
it provides benefits that are not available from Microsoft.

I hope this helps answer your questions.  And thank you very
much for the invitation to send your our press release when it's
ready.

Greetings, Norbert.

> Hi, David, Norbert,
> 
> I'm writing a story for LWN.net on dotGNU (and Mono, of course) for this
> Thursday.  I've been reading over the stuff on the web site and mailing
> list, and I have one question that I can't answer for myself:  why a new
> bytecode language and virtual machine?  It seems like a tremendous amount
> of work that duplicates things that already exist.  Is there a reason why
> you choose not to (1) use the common language stuff that Mono will develop,
> or (2) start with something like Java, for which free VM implementations
> already exist?
> 
> If I were trying to start a DotGNU, I would probably accept the .NET-style
> plumbing, as done by Ximian, as a given and work on the stuff that really
> matters, like the virtual identities.  *That* is something I would like to
> see take hold.  I'm not doing this, of course, but I would like to
> understand your reasoning.
> 
> Please feel free to drop us a copy of your PR once you've got it into
> condition for publishing; we'll be glad to put it out there.
> 
> Thanks,
> 
> jon
> 
> Jonathan Corbet
> Executive editor, LWN.net
> address@hidden
>