improve zip/archiving of download files #2920
Description
A user reported seeing CRC unzip values errors/warnings; for example: "extracting: BIOM/57457/all.biom bad CRC f6b2a86b (should be 38903659)".
Background: one of the largest changes from last release was that now we are storing the file size and it's CRC32 checksum in the database; this checksum is added to mod_zip/nginx so it can be used for file download resume. To be clear, the CRC of the file (f6b2a86b) is the value stored in the DB, displayed in the Qiita pages; and the value you get in a laptop once you download the file via the crc32 command AKA this value is the correct one. Also note, the downloaded zip contains the correct files; even if zip complains, the files are fine. Thus, there is something "odd" with mod_zip/nginx.
After further investigation @ElDeveloper and I realized that:
- mod_zip doesn't actually compress the files (it creates a zip file without compression) and it's part of their design: Zips are not compressed evanmiller/mod_zip#23
- the checksums added to the zip file are incorrect