Duffbert's Random Musings

1 - Tom, a comment/thought relating to your 'Check for Response block' agent... at the moment, if you don't have a match, you end up doing five view lookups per comment. If you're not expecting too many blocked posts, then why not flip that around, and do the checks based on the first parts of the IP address, progressively doing more detailed checks?

If the first octet (say) doesn't match, you know you won't have a match at all and can exit without having to do the rest of the lookups. That way if your most common case is to accept a post, then you incur less overhead, and you only pay more for actual matches.

(I know that in a low-traffic situation it won't make much difference... but I'm just mulling over the best way to handle IP lookups/matching in other potentially higher-traffic situations...)

2 - I like that... the "urgh" factor...

When you first suggested it, I was thinking I'd have to have three additional views that showed IP addresses by 1st octet, 1st/2nd octet, and 1st/2nd/3rd octet. I'd then do a search on the 1st octet view to see if I matched the IP of the response to any block records. If no match, you're done. If you do have a match on the first octet, then you'd search on 1st/2nd for a match on either the first two octets of the IP or the first octet followed by an asterisk... Repeat as necessary...

I'm lazy... I like my method better...

3 - Ben: Thank you very much...

Colin: Welcome to the blog... You make a good point there. Since I was doing a lookup to actual entries in the blocked address view, I didn't break out the address by octet for lookup purposes. I'll have to rethink that idea and see what I come up with. Thanks for the input.

4 - Thanks Tom. The more I think about it the trickier it seems to be though, since at each potential 'match' you'd need to decide whether you've found a wildcard match, or a partial hit where you need to keep looking. Hrrm, that probably means more I/O overhead... I'm not sure whether that negates any savings in lookups, not to mention the urgh factor in the code.

A pre-built document/data structure that gets updated each time a new IP address/range is added might be more efficient, but the complexity is even higher.

Further to Ben's recent optimisation thread, I wonder if the additional effort would really be worth it. I know it's going to bug me silly, though

5 - Good work Duffbert! I had to do something similar with referrers and searches some time back on my site, and have been meaning to extend the IP address, host and content checking to the comments process.