*  Doing a HEAD on an ISMAP link causes a 500 server error. Maybe a 
   GET will be better for this.

*  Add a mechanism for entering authentication data in order to check
   such areas as well. Possibly allow multiple pairs. Rather not enter
   the passwords on the commandline.

*  Obey Robot rules. My current idea is to obey robot rules by default
   on all 'external' request, and ignore them on 'local'
   requests. Both of these should be changeable.

*  Make it possible to convert http:// references to file://
   references, so that a local server can be checked without going
   through the WWW server.

*  Keep state between runs, but make sure we still are able to run
   Checkbot on several areas (concurrently). Uses for state
   information: list of consistent bad host, remembering previous bad
   links and just check those with a `quick' option, report on hosts
   which keep timing out.

*  Retry time-out problem links after checking all other links to
   better deal with transient problems. See above as well.

*  Add problem links such as the 500's series together.

*  Parse client-side (and server-side) MAP's if possible.

*  Maybe use a Netscape feature to open problem links in a new
   browser, so that the problem links page remains visible and
   available. Frames? (*shudder*)

*  Include (or link to) a page which contains explanations for the
   different error messages. (But watch out for server-specific
   messages, if any)

*  The external link count is way off. Write code to parse the
   external queue first, and then run through it to actually check the
   links. 

*  Add an option to indicate which error codes can be ignored by
   Checkbot's reports.

*  Keep an internal list of hosts to which we cannot connect, so that
   we avoid being stalled a while for each link to that host.

*  Report all pages to point to a particular broken link. Most of this
   information is already availabe, but discarded when writing the
   report pages. Should probably change the internal structure a bit
   to get this working properly.

*  Rewrite report generation. Currently the whole error list each
   parsed for each host for which we write a file. This can become
   very time-consuming!

*  The exclude option is somewhat confusing, in that the links
   matching this option will still be checked (the links on these pages
   will really be excluded). Maybe add a new option to really ignore
   options? Or just redo current options.

*  Add an option to count hops instead of using match, and only hop
   that many links away? Suggested for single page checking, but might
   be useful on a larger scale as well?

*  Sort problems on the server page in a different order
   (e.g. critical errors first).

*  Improve the reporting pages to be more clear about what is going
   on, etc.

*  Using IP address in reports can conflict with HTTP 1.1 virtual
   hosting
