At 11:08 15/06/2006, Denis Walker wrote:
Tim Streater wrote:
At 10:52 15/06/2006, Marco Hogewoning wrote:
On Wed, Jun 14, 2006 at 12:23:54PM +0100, Niall O'Reilly wrote:
Begin forwarded message (extracted from a recent update attempt):
changed: Niall.oReilly@ucd.ie 20060614 changed: oonagh.oreilly@comreg.ie 20040323 changed: hostmaster@ripe.net 20041222 source: RIPE ***Error: The dates in the changed: attributes should be in accending order '20040323' was found after '20060614'
Wouldn't it be just as easy to sort the attributes into the correct order as to generate an error message?
Agreed, and as I also just got bitten by this. The strange thing is that at it also reports 'syntax check passed' first and then spits out the error. Isn't this a syntax error ?
Nothing wrong with the syntax of the statements. It's a semantic matter - there is meaning in the content of the statements that implies an order. However it would seem reasonable that this be fixed up automatically.
We can easily 'fix' this by re-arranging the order of the changed attributes. But the convention has always been that the software does little or no changes to users data. If the semantics are wrong, we report it and let you fix it. There are many simple errors like this the software can fix. But once we go down this road we run the risk of automatically fixing problems in a way that the user did not want.
This changed attribute date is one example. This error may be because the user put a new changed attribute with the current date before a previous changed attribute. Or it may be because the new changed attribute had the wrong date, but was in the correct place. In the latter case, if the software re-orders the attributes, you will not realise that you have used a date of 200501 instead of 200601. A very common mistake in the new year.
Yes, that makes good sense. -- Tim