It looks like these are things like smart quotes (generated by Word, amongst other things). The only thing you can really do is identify the exact binary representation of the character in the offending web page, and do a search and replace using that.

Also, you need to replace the longer strings before the shorter ones in case the the shorter string appears within the longer one - that will really muck things up...

It may be possible to fix this by switching the page character set but it hasn't worked for me so far in my tests.




This one is easy, by the look of it...

Each HTML page has a {body} tag in it - this needs to be changed to {body bgcolor="black"}.

(The braces should be angle brackets, but they won't render properly, and I can't remember how to force it)

When doing the search, make sure the strings are specified with both open and closing angle brackets just in case the body tag already has other elements (which will usually include a bgcolor element)




Depends on how you are doing the search and replace, and exactly what is going wrong with the working pages. If it is a straight file-age-related thing then you could copy the offending files to a working directory, update them, then copy them back.

As noted, you may also be hitting the case where some of the odd strings are subsets of the other - for example if one code was ABCD and the other was ABC and you needed to map ABCD to WXYZ and ABC to MNO, you would have to remap ABCD first. If you remap ABC first you would end up with a lot of MNODs.