Re: [Fresco-devel] Server-resolution/publishing algorithms and proposal
Nick Lewycky <[email protected]>
| Newsgroups | gmane.comp.video.fresco.devel |
|---|---|
| Message-ID | <[email protected]> |
Stefan Seefeld wrote: > graydon hoare wrote: > >> I don't usually comment much on goings on here anymore, but this is an >> issue I remember having a suggestion for: forget about IORs and name >> service. use an environment variable and make it just a tcp/ip >> "host:port" pair. when you connect to that port, it tells the client >> whatever else the client needs to know (eg. an IOR for the server root, >> or whatever). > > Who is 'it' ? It sounds like a (degenerated) name service in any case. A degenerated name server minus the annoying hangs that omniORB's nameserver is so famed for. There's no reason we couldn't implement a light-weight server as one of the export-ref options now and decide whether or not to make it default later. (Or even enabled irrespective of the export method.) At the moment, we just cram the IOR straight into the environment. What I don't see is why making the env hold a host:port pair to find the IOR is any better. From the point of view of the user, it would be the difference between: $ FRESCO_DISPLAY=host:port; bin/demo vs. $ bin/demo -R lookup -i host:port if we were to tie it into the -R option. Nick Lewycky