Doc Shipley wrote:
Jules Richardson wrote:
Jim Leonard wrote:
CPU speed and blank DVDrs are cheap enough
nowadays that this adds 1%
of effort to a process that gives me 3-4x more reliability.
... until you find a few years down the line that your burner was
producing CDs that refuse to work in any of the hardware you can now
lay your hands on.
As I said a couple of weeks ago, there's *no* archival media that's
exempt from that kind of obsolescence.
Absolutely. What I'm getting at is CDs seem to be at the very low end of the
scale in terms of compatibility - conversely things like tape, floppy, SCSI /
ATAPI devices etc. seem to have far less chance of interoperability problems.
Can't comment on DVD; I don't use it for anything.
I could take a SCSI disk and plug it into any old SCSI HBA, or an IDE disk on
an ATAPI controller, or a 3.5" floppy in a 3.5" drive, or a DLT drive on a
SCSI HBA etc. and it's *almost* certain that the system could access the media.
With CDROM, those chances are a *lot* less when all possible drives are taken
into account (laptop drives and 'combo' drives that read both CDROM and DVDROM
seem to be notoriously bad at reading non-commercial media)
So you're right, refreshing often is the sensible thing to do - but despite
that I'd still rather choose media that has a high level of compatibility
across hardware / systems than one that doesn't, just in case.
Archive it, TEST IT, duplicate it on different
media, test *that*, and
migrate often.
Yep. But testing - whilst obviously necessary - still can't of course provide
any guarantees that the media will work on whatever hardware can be sourced at
the time of needing to access it. The best anyone can do is use something
that's *reasonably* likely to work, but there are no certainties.
Of course knowing that there will be a backup prog around down the line
that'll understand the contents of your media is another matter entirely :-)
(which I suppose is why simple formats like tar are so useful - even if tar
vanished off the face of the map, the format can really be inferred from just
inspecting the data and it's trivial to knock together a reader)
cheers
Jules