8 Nov
2021
8 Nov
'21
4:44 p.m.
That's an excellent argument.? I was thinking that knowing the OS type
might help the user not issue commands that the remote system couldn't
do.? Tops-10 file names were what came to mind.
However, feature flags clearly could handle this.
The only thing I can imagine is a 'what if' scenario where somebody
dreams up something that the feature flag doesn't /currently/ cover; the
OS Type could be used as a heuristic until the feature flag was defined
and the implementations caught up.
That would only be a stopgap measure.? Special casing by OS type can get
you into trouble if you then have to start worrying about versions of
OS's...
------------------------------------------------------------------------
On 11/8/21 9:57 AM, Paul Koning wrote:
I don't know why DAP includes an OS type code at all. When I was working on DECnet
architecture I repeatedly argued that DAP implementations that do OS type checks are
broken, because they are assuming that they can correctly deduce the properties of some
remote system from its OS code. The correct way -- which is supported, in detail, by DAP
-- is to look at the feature flags to determine what the OS can and cannot do. And if
those flags are insufficient for a particular case, the solution is to extend the protocol
to add the missing information.
In other words, a client should conclude "the server can do X because it told me it
supports X" and not "the server can do X because it is running OS Y".
paul
------------------------------------------------------------------------
On Nov 7, 2021, at 6:40 PM, Thomas DeBellis<tommytimesharing at gmail.com> wrote:
...
On the other hand, I don't remember my FAL/NFT/DAP blowing up on the wrong system,
either (they blow up on plenty of other stuff...) It would appear that there are two (2)
separate lists of system type bytes, one for NRT and the other for DAP, viz:
...