I understand this is the least of our worries right now, but I'd like to get the indexer cache working so I can get karma for PhD. :-D
The main troubles with serialize is that it 1. wastes space by not preserving internal references and 2. requires lots of memory. I recall SQLite being proposed as a possible solution, but since the entire index is loaded into memory I don't think that's necessary: just as easily parseable file format. So, we have a file, with each id seperated by a newline, and the fields seperated by tabs (both of which should never occur in the fields, if they do I'll use a control character or something). children is collapsed into a list of IDs. Reassembling is as simple parsing the file line by line, constructing the ID array from the fields. Then the children are reconstituted by running through $IDs a second time, replacing them with references to the appropriate indexes. Thoughts? -- Edward Z. Yang GnuPG: 0x869C48DA HTML Purifier <http://htmlpurifier.org> Anti-XSS Filter [[ 3FA8 E9A9 7385 B691 A6FC B3CB A933 BE7D 869C 48DA ]]