I believe ZIP supports updates about as well as SQLite. You append the new file to the end (and write the "header" after that). This leaves a hole in the middle, but that can be reused by a later addition. There's some details about internal page size and fragmentation, but I think it could amortize to about the same. There's something to be said about quality of implementation, but that's not an argument about the file format per se.
How? There's a massive difference in usability and performance with a SQL database and all the data querying and manipulation it provides compared to a ZIP file that just straps together a bunch of compressed payloads with a key/value file structure.
The linked page even has a case study of the OpenDocument file format that shows exactly how SQLite is better than a Zip file: https://www.sqlite.org/affcase1.html
If you're using SQLite as a pile of files, you're not doing much querying if the compressed data inside.
> Newer machines are faster, but it is still bothersome that changing a single character in a 50 megabyte presentation causes one to burn through 50 megabytes of the finite write life on the SSD.
That's funny, since Firefox would burn through gigabytes of writing to SQLite. All comes down to quality of implementation.
The point is that you don't just use it as a pile-of-files when you have SQL access to the data and instead use a high-level data model. This is not possible at all with ZIP.
I honestly don't know much about the relative differences in formats like this, what is horrible about ZIP? Also, my impression is that more compressed files end up in RAR these days. I know that's not a new format either-- is it similarly horrible?
The central directory is a great feature of zip. Makes it easy to append to and to list the files or decompress individual files without decompressing the entire archive.
The actual horrible thing about zip is that there is no standard whatsoever for the encoding of filenames. So mangled filenames are common and most utilities don't even let you manually select the encoding if you know what it is.
Are there zip libraries that facilitate this usage pattern?
Either way, I tend to think you'd probably be better off with sqlite since it would give you more flexibility (for instance if you later realize that your application possesses relational data.)
