Gossamer Forum
Home : Products : Gossamer Links : Version 1.x :

Verifying Links

Quote Reply
Verifying Links
Alex:

Code:
That's a perl 5.6 issue, we had a typo in nph-verify.cgi:

use vars qw/%ok_status %fail_status %all_codes/;

make sure you remove commas between the ok_status, fail_status etc.

This worked.

When verifying links, because of the size of the database, there is a "size" problem with the ODP import.

Some way of randomly assigning a checked date to them is needed. How about something that would start with link 1, and check every 7th link one day. Then start with link 2 and check every 7th link, etc.

That would check 1/7th of the database at a time, and give some initial breakdown of the database.

Also, how about an option when the error message:

Checked 842 - Success (302). Message: Found, but data resides under different URL (add a /)

comes up, the URL is automatically adjusted and marked as good? This could be an option in the run command to "fix URL's when possible" or something like that. The current "repair" feature doesn't seem to do this quite right, and I've gotten URL's with double // and filename.html/ from the repairs.

In order to do this correctly, the URL's would probably have to be compared after the redirect message, to make sure it really was a missing '/' and not some other reason.





------------------
POSTCARDS.COM -- Everything Postcards on the Internet www.postcards.com
LinkSQL FAQ: www.postcards.com/FAQ/LinkSQL/











[This message has been edited by pugdog (edited May 09, 2000).]
Quote Reply
Re: Verifying Links In reply to
Also,

One nice feature would be to move all "dead" or "missing" links to an "on-hold" database, and allow checking and re-activation of those links if/when they come back alive. Sometimes you want to archive dead links, at least for a short time for all sorts of reasons.

I've made some changes to the link selection, especially for "popular" and "new" links (it's suprising how many new links die in days), such that if status =404 or <0 then the link is ignored for listings.

On a build, I've used AND Status='202' to pick links out, because it's a faster test than a not and/or multiple selection.

But, if I actually can move the links OUT of the database to a "hold" database, then I don't have to do any of this link testing during builds.