> On Jan 17, 2023, at 10:58 AM, Kenneth Gober via cctalk
> <cctalk@classiccmp.org> wrote:
>
> ...
> There is another significant advantage to focusing on a backup strategy as
> opposed to an archival strategy -- I don't need to worry about tapes I made
> today being readable decades in the future (i.e. I don't need to worry
> about keeping legacy tape drives running). The file server is the
> "archive" and I only *need* to be able to read the tapes I made last year
> or last month. If a tape drive fails I can just get another one, and if
> they become hard to get there's nothing stopping me from upgrading to
> something newer with better availability (i.e. I don't have a library of
> old media that I need to be able to read forever). The disadvantage of
> course is that a file server isn't something you can reasonably put in
> storage for future generations to access. This strategy really only works
> if the file server is accessed frequently so that
> accessibility/compatibility issues can be fixed incrementally as they
> happen (while they are still easily addressed using current/recent
> technology).
>
> -ken
Nice idea and a very good point. You have to deal with a number of
compatibility issues but they are all short range. The backups is one aspect.
Another is the file server protocol. If your server uses NFS V2, some hosts
can use it but others may not. And the hardware on which it runs might fail.
For these last two cases you'd have to stand up a newer file server, and in the
case of server failure you'd have to be able to load the new server with the
old data. Depending on the backup media that might be problematic. I suppose
for both of those issues the answer is to bring up newer file servers ahead of
time, as they become available.
You still have to deal with the question of how old format documents would be
read. If the formats are proprietary and no longer supported that's a problem.
For example, I have some old Eagle CAD files that are no longer readable. And
some old CorelDraw files that are also difficult, though I think I can still
get there via intermediate old versions of the software on a VM. And I have an
old CPLD designed in Abel HDL; if that ever needs modification I'm going to be
in trouble.
paul
