VMS & TDS: a summary so far

[Philip Taylor (RHBNC)]

Karl --

>>     I can't agree with this: speed is purely a function of the processor,
>> 
>> I believe the speed of the disk is the limiting factor.

True: I over-simplified.

>>     what do non-VMS TDS systems do?

>> As Sebastian said, web2c supports a file (ls-R) that contains a recursive
>> directory listing of all files in the tree. ls-R is read at the first
>> reference and serves as the ``cache''.

>> In the default configuration, if a file is not found in the cache, it is
>> then looked for on disk. Thus, if you install a new file in the tree, it
>> will be found (just very slowly, by comparison). 

A good algorithm: I like that.

>> (Many people disable
>> this, though, and use only the cache file -- their choice.)

Hmmm...

>> If on the other hand you install a new file in the tree and an old file
>> by the same name (that can be found along the same paths) still exists
>> in the cache file and on the disk, then the old file will be
>> found. That's the price you pay.

Not so good...

>>     is a compliant TeX required to do the same thing for TFM files?).
>> 
>> I don't understand what you mean. Caching is not required for TDS
>> compliance, because when we asked all the developers we could find, most
>> of them didn't want to commit to implementing it.

I wasn't referring to cacheing but to tree-searching: is a TDS-compliant
TeX required to tree-search for TeX files, TFM files _and_ format files?

>>     I want to be able to edit a file, create a new version, and _know_
>>     TeX will find the new version the very next time it runs.
>> 
>> In that case, you obviously can't have a simple cache file as outlined
>> above. Your call.

It will be interesting to see whether the VMS file system cache could be
tuned to provide the functionality and performance that we need.

** Phil.