I'm unclear on how SQLite (native format, no zip) is achieving sizes similar to XML + Zip. Are SQLite TEXT or BLOB fields compressed? Or are they assuming the caller is compressing BLOBs before writing?
Engineering is all about tradeoffs: SQLite is optimized for quick incremental updates where you don't need to rewrite the whole file. Zip & xml aren't. (IE, if you decide to add a letter to a word at the beginning of a document, with zip & XML you have to rewrite the whole document. SQLite can make a minor change without the whole rewrite.)
In our case, file size was not a factor in choosing between SQLite and XML.
But, remember that file size is deceptive: Disks are block devices; the 30 byte and 1k file take up the same space if you block size is 2k. (I've shipped a filesystem driver.) HTTP servers gzip on download. It's more important to know your needs than to get hung up on a single metric like file size.
> I'm unclear on how SQLite (native format, no zip) is achieving sizes similar to XML + Zip. Are SQLite TEXT or BLOB fields compressed? Or are they assuming the caller is compressing BLOBs before writing?
Remember, XML writes each tag name 1 time if there's no content and twice if there is. Each attribute has it's name written every time. I doubt SQLite writes all the metadata in each row.