[db-wg] changed-lines in objects.

On 2003-05-24 16:25:38 +0930, Mark Prior wrote:
> At 8:12 AM +0200 24/5/03, Fredrik Widell wrote:
> >
> >A thought occured to me (as I for the xxxx'th time was accused of
> >hacking somebody who has a firewall). The changed-line in every
> >object simply refers to a mail-adress and a timestamp. Whould'nt it
> >be more accurate to have a nic-hdl instead of just a mailadress
> >there? managers of objects move, and mailadresses gets obsolete,
> >and the only thing they are used for is to send abuse-matters,
> >which of course are the wrong recipient in 99.9% of the cases, but
> >if there was a nic-hdl, one could give information via that nic-hdl
> >on how to get in contact with 'good' recipients, or the users would
> >have to actually think and maybe realize that the other names that
> >occur in an answer from ripe-db possibly could be the right
> >recipients for abuse-matters.
> >
> I think they are just about useless anyway and anything you put in
> there some clown (or their program) will use to send you either SPAM
> or abuse about SPAM. The objects I modify usually have a comment
> about where to send abuse reports and the changed field has
> nobody at domain as the address, which nicely hits the bit bucket for
> those people who can't read/understand.

At the Database Working Group of the RIPE meeting last week, the topic
of the e-mail addresses in the "changed:" attribute came up.  I asked
the participants how many people actually used the information in the
"changed:" attribute for themselves to track who had made changes in
their objects, and a significant number said they did.  This was
surprising to me, but means that people do find it useful.

I was talking with one of the engineers in our group here, and he made
the suggestion that we should make the changed attribute:

changed: YYYYMMDD <freeform>

When an object is updated, the value of the "changed:" attribute would
be checked, and if the first 8 characters were not the date, then the
date would be inserted.

This attribute would preserve the useful features of the "changed:"
attribute:

- automatic timestamping for users who want it
- place to put an identifier for the person making the change

It would solve the problem of having a meaningless e-mail in the
attribute.

A further improvement would be to make the attribute optional, since
it has no strict meaning.

It does not address the problem that people assume that the "changed:"
attribute actually properly identifies when an object was changed.  If
this is desired, then the attribute should be made mandatory, and the
timestamp always added by the server.  In the past, there has been
opposition to this idea.

-- 
Shane Kerr
RIPE NCC