3 Nov
2010
3 Nov
'10
3:14 p.m.
On 3 Nov 2010 at 20:30, Tony Duell wrote:
I am seriosulyt worried by the fact that a group of
programmers who
are capable of writing a compiler didn't realise the value of backups.
I can't beklieve tht nobody had ever lost data before.
It wasn't as bad as it sounds. Since implementation was "divide and
conquer", copies of the source existed in some form among the members
of the implemenation time. Executables were already widely
distributed around the company and to a few customers.
Had we lost the source entirely, the practical upshot would have been
that someone would have had to spend a day recreating the source
disks from team members' copies.
We all had years in mainframe product development and were well aware
of the fragility of electronic storage. That there was only one set
containing the collected source would have mattered little.
We'd come from an environment where you started a session with *no*
files. At session end, any files created during the session were
disposed of--certain names, such as OUTPUT and PUNCH were sent to
their appropriate queues, while any other files were discarded.
If you wanted to use a previously-created file or create a new
"permanent" file, you had to specify its catalog name,version (you
could have multiple versions of the same file), account and password,
session permissions and the local file name (i.e., the local
association) for that file.
Backups were handled by system operations on a daily basis, so
individual users never had to worry about making personal backups.
Inadvertently deleting or modifying a "permanent" file was almost
completely unknown.
In comparison, Unix seems positively crude in the respect of data
security (one aspect of which is to protect a user from himself).
And micro-oriented operating systems such as ISIS or DOS are
downright primitive.
--Chuck