TXT2HTML TODO LIST Please Suggest More Improvements Notes by David Mertz mertz@gnosis.cx Tue 05-16-2000 PROGRAM STRUCTURE IMPROVEMENTS ------------------------------------------------------------------------ - Should the whole source present one or a few classes with methods? It might be better to eliminate cfg_dict and other global-ish variables, and make them instance data? But is there a real gain in such a change? - How might the elements that change be better encapsulated: regexes for new types, case blocks for handling of types, options in proxy form, type inference? - Is it worth going to several module files? A package? DOCUMENTATION IMPROVEMENTS ------------------------------------------------------------------------ - Write a HOWTO document for adding new functionality? OUTPUT IMPROVEMENTS ------------------------------------------------------------------------ - Syntax highlighting for other languages. Are there other tools that would be useful to call for various languages? GNU enscript, maybe (but seems to rely on many installation assumptions). Is it worth wrapping a non-Python tool? - FAQ TYPE conversion basically *sucks*. It works pretty well for the comp.compression FAQ that served as an example; but it looks terrible on every other example encountered. Is there any consistent pattern in FAQ documents? - An "EMAIL" TYPE? What would it do, exactly? - A "Project Gutenburg" type. - What about the dependency on the external 'eye.gif' icon in the proxy navigation box? How might that degrade more gracefully if icon is missing? (PIL!?) USAGE IMPROVEMENTS ------------------------------------------------------------------------ - Some sort of mechanism for "bookmarking" URL's that are actually improved by being proxied. Crude fixed list 'nicelinks.txt' has been done for the concept. What interface to store bookmarks? How to handle users, permissions, etc. (or do not bother)? Should 'nicelinks' get its own TYPE while we are at it? - Some links should never be proxied, such as .zip archives, or fancy types that use plug-ins. Basically, anything that isn't more-or-less ASCII to start with won't be handled right. How to assure this? Can we enumerate all the types on either side (probably not)? SECURITY ISSUES ------------------------------------------------------------------------ - Are there any weaknesses in use of CGI proxy? Attacks on my host? Spoofing of HTTP_REFERRER for some bad purpose? Something else?