Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!wuarchive!usc!sdd.hp.com!decwrl!sgi!shinobu!fido.wpd.sgi.com!marktwain.rad.sgi.com!linton
From: linton@marktwain.rad.sgi.com (Mark Linton)
Newsgroups: comp.windows.interviews
Subject: Re: Basis for RPC?
Message-ID: <1991May6.184950.2195@fido.wpd.sgi.com>
Date: 6 May 91 18:49:50 GMT
References: <STEINARB.91May4124947@fenris.idt.unit.no>
Sender: news@fido.wpd.sgi.com (Usenet News Admin)
Reply-To: linton@marktwain.rad.sgi.com (Mark Linton)
Distribution: comp
Organization: sgi
Lines: 17

In article <STEINARB.91May4124947@fenris.idt.unit.no>, steinarb@idt.unit.no (Steinar Bang) writes:
|> I saw some comments to the effect that C++ streams are out as a basis for
|> InterViews RPC (or rather: will be removed when time permits).
|> Instead memory mapped files will be used on the systems that implements
|> them.
|> 
|> I can see that this is more efficient than using streams. Two things that
|> bother me, however:
|>  - Not all systems will implement memorymapped files.
|>  - Net transparancy will be lost.
|> 
|> Will you continue to support RPC classes that uses streams (maybe based on
|> ANSI C <stdlib> function calls)?

My comment was in reference to file i/o, not rpc.  The idea is to have a File class
that hides allows mmap to be used; it could also support normal read/write.
I don't expect this to have any impact on rpcstreams.