[SCC_Active_Members] Software Archive Solution

[Richard P. Gabriel]

At 23:35 -0700 4/10/07, Bob Fraley wrote:
>There is a difference between my proposal and the examples given 
>here.  In the examples some media has been written and left 
>unattended for years.  In my proposal, the master archive is active, 
>kept in an operating computer.  So if storage technology changes, 
>the archive will need to be copied to the new media.  Of course, if 
>the archive were ever to be taken off line, the backup media would 
>become the archive.  By doing periodic full backups, the technology 
>for storing the backups will be recent technology.  The older 
>artifacts will be available on recording media that is contemporary 
>with this event, rather than media contemporary with the artifact 
>itself.

I was assuming the archives were on live machines and constantly and 
instantaneously available. Why would anyone assume otherwise?

>When new file formats are created, an automatic process could go 
>through the archive to generate the new file formats.  Similarly, 
>audio or graphics files could be converted to new formats while the 
>conversion programs are available.  The archives that you reference 
>were allowed to sit idle until the formats were no longer in use, 
>and any conversion programs that may have existed were no longer 
>available.

Well, you make a rosy assumption. There never were any conversion 
programs for SAIL. "Standards" grew up separately (essentially in 
silos), and the silo that was SAIL was merely abandoned. The work to 
make the SAIL files available started very soon after the machine was 
decommissioned, by what were then called system programmers (today 
maybe called senior sys admins - there is no parallel because the 
people back then actually were usually the OS implementors or near 
kin; imagine a Linux committer running a Linux farm), and they wrote 
the first conversion programs.

Step by step conversion is not always a reasonable alternative. 
Consider the SAIL case. The SAIL character set included several 
mathematical symbols, alpha, beta, for each, there exists, union, 
intersection, etc, etc. When the conversion was done, unicode was not 
a real alternative, so they used an html encoding that captured these 
characters, but they were custom encodings. Now one could convert to 
unicode, but the archive that is on a live machine is the transcoded 
one, and so another custom conversion program needs to be written.

You might argue that SAIL is just a backwater and who cares, but for 
20-30 years it was among the top half-dozen research labs in the 
world.

What I'm saying is that it is unlikely to be the case that a routine 
process can be followed.

>  Clearly, the curators will need to remain alert to apply the 
>appropriate translations in a timely manner.  Formats that are no 
>longer in use can be removed from the archive, but the core data (in 
>the 1620 case the binary record of all holes in the card) should 
>never be removed.

I sure wouldn't delete these intermediate steps. Otherwise you might 
not be able to fix mistakes with reasonable effort.

My sense is that the OAIS model has examined most of these concerns 
I've brought up.

			-rpg-