Applications Messaging Standard

[Tony Linde]

 > I think even the suggestion of using the Windows registry violates
 > the first comment above (can somebody point me to a Fortran
 > interface to the registry?).   As has been pointed out before, the
 > current .plastic file (and the suggested .ivoamsg file) is not the
 > reason for your demo problems and so long as apps play nice with
 > the host system, the use of dot files is not ground-breaking technology.
 > Likewise, using the java properties is nice for java, but....

I wasn't suggesting that we use the windows registry - I said that the registry is 
the obvious solution in windows but that since linux does not have an equivalent 
facility (don't know about macs), the common file seems to be the lowest common 
denominator.

One point since we're mentioning languages: do other languages have the concept of a 
'home' directory? We know that it'll work for java and .net languages, but do 
python, perl etc have commands to get the 'home' directory which will return the 
same directory as java does?

 > the first comment above (can somebody point me to a Fortran
 > interface to the registry?).   As has been pointed out before, the

If there is a windows Fortran compiler then it'll have access to the win32 libraries 
which make registry access relatively simple.

T.

Mike Fitzpatrick wrote:
> On 2/15/07, Tony Linde <Tony.Linde at leicester.ac.uk> wrote:
>> Hi John,
>>
>>  > whatever solution we choose should be platform and
>>  > language neutral.
>>
>> agreed.
>>
>>  > Unfortunately, there's no way
>>  > for Python, Perl and everyone else to use it.
>>
>> I'd be surprised if there weren't libraries for getting the windows
>> registry info - not sure what java does so don't know about that.
> 
> I think even the suggestion of using the Windows registry violates
> the first comment above (can somebody point me to a Fortran
> interface to the registry?).   As has been pointed out before, the
> current .plastic file (and the suggested .ivoamsg file) is not the
> reason for your demo problems and so long as apps play nice with
> the host system, the use of dot files is not ground-breaking technology.
> Likewise, using the java properties is nice for java, but....
> 
> I also think we're agreed that a well-known file, a separate name server, a
> "hub" or somesuch is needed and is an implementation detail.  For this
> exercise we should concentrate on what information needs to be written
> to establish the connection and pass messages.  Does the XPA name
> server do something the PLASTIC hub doesn't?  If so, do we need it and
> what does that look like in our new protocol (i.e. a dedicated 
> administrative
> message or some generic functionality) ?
> 
> -Mike