20 Dec
2019
20 Dec
'19
7:33 p.m.
On Dec 20, 2019, at 1:12 PM, Thomas DeBellis
<tommytimesharing at gmail.com> wrote:
I have been working with a kind volunteer who is using OpenVMS Alpha. It's been
quite productive.
While I haven't run into any security exposures, yet, the experience has enabled me
to find a few things to tweak up. For example, the VMS logging and reporting was clearly
superior to what was being put in the FAL log (which wasn't even timestamped and
didn't have the requested file name).
We ran into some interesting behavior. When ANONYMOUS access is restricted to the
stand-in user and its subdirectories, wildcard listing requests that are outside of that
lexical scope do not report any errors but rather return zero files (none). That was
puzzling because the part of the code that did the access check was clearly doing the
reject with appropriate return values.
Not clear if that's an error. In RSTS at least, if you asked for a list of files,
you'd be given the subset of files that you have read access to. Files for which you
have no permission would appear nonexistent. This is rather different from the behavior
in Unix, which shows you all files in a brief listing, and a pile of error messages in a
long listing.
Whether VMS follows the same model I'm not sure. It's a bit more work in systems
like RSX or VMS, which like Unix separate directory names from the underlying file object
where the permissions live.
Further analysis showed that in the main loop for
stepping wildcards, a reject causes the file in question to be skipped; nothing is
actually done with the return values. I believe this may be intended behavior for two
reasons. The first is that if the wildcard includes files to which the ANONYMOUS user
does legitimately have access, it's friendlier to report those as opposed to stopping
everything.
The other reason I believe has to do with DAP itself; I don't think you can't
return more than one return value per ACCESS request, even with a wildcarded file
specification. So perhaps this isn't a bug so much as an unintended consequence?
Does anybody know what the behavior should be?
At any rate, I don't believe the current behavior is 'informative'. What
I'm thinking of doing is saving the last error code and returning that if the count of
files returned in list is zero. Otherwise, the user had something to look at.
Do I have any volunteers who operate something besides VMS? I'd be interested in
RSX(+), IAS, RSTS and OS-8. They never did DECnet on RT-11, did they?
I can offer help with RSTS. Yes, there is DECnet/RT11, but I don't know if we have
any on HECnet or if the code has even been preserved.
paul