Zip Files

The numbers above are the ones that usually decide how an archive should be built. Each file you add can be up to 95 MB, and the finished archive is removed from our servers 30 minutes after it is made. On the receiving end, email is the tightest limit: Gmail and Outlook both refuse attachments past a set size, and they count the encoded message, which is noticeably larger than the file you see on disk because attachments are encoded as text for the trip.

This is why archives are often built to a size rather than just packed. When the result has to travel by email, set Split into parts of to a value under the mail limit and send the parts in separate messages, or upload the single archive to a cloud drive and share a link instead. The 4 GB line only matters for very large single files such as disk images and video masters; modern tools write ZIP64 automatically when a file passes it, but some old extractors on embedded devices and older systems still cannot read ZIP64 entries.

Compression level changes how hard the packer works, not what you get back. Every level restores the files exactly as they were. Store packs without compressing at all, which is the quickest choice for photos and videos that will barely get smaller anyway. Fast and Normal suit everyday mixes of documents, and Maximum spends more time searching for repeated data, which helps most on large text files, logs, databases and source code.

It helps to know in advance which files will get smaller. Plain text, CSV exports, spreadsheets saved as XML, log files, uncompressed WAV audio and BMP or TIFF images often pack down a lot, because they contain long runs of repeated data. JPG photos, MP4 and MOV videos, MP3 and AAC music, and most PDFs are already compressed by their own formats, so a ZIP of them comes out close to the original size. If the goal is a smaller file rather than one bundle, compress the media itself first with a format-specific tool and then zip the results.

ZIP has been the common denominator since Phil Katz published it at PKWARE in 1989, and it is still the format a stranger is most likely to open with a double click. Each file inside is compressed on its own, usually with the Deflate method described in RFC 1951, which is why a damaged ZIP often still gives back most of its files.

7Z, from the 7-Zip project, compresses with LZMA or LZMA2 and can treat many files as one continuous stream, a technique 7-Zip calls solid compression. That usually produces a smaller archive when the files are similar, such as a folder of text or code. The 7z format page lists AES-256 encryption and compressed headers among its features, which is what lets it hide file names.

TAR.GZ is two steps in one: tar joins the files and keeps Unix permissions and folder structure, then gzip compresses the whole bundle. It is what most Linux software and server backups use, but it has no password option of its own. If a TAR.GZ arrives and you need it as ZIP, the archive converter page repacks it.

ZIP supports two kinds of password protection. ZipCrypto is the original scheme from the early ZIP days and has well-known weaknesses, so treat it as a lock against casual opening rather than real protection. AES encryption for ZIP was added later, following the format WinZip documents for AES archives, and is the default here. The catch is compatibility: some built-in extractors cannot open AES ZIPs and ask for an archive app instead, so tell the receiver in advance.

Send the password through a different channel than the archive, for example a message or a phone call when the archive goes by email. A long passphrase of several unrelated words is easier to say aloud and harder to guess than a short string of symbols. Keep an unencrypted copy of anything you cannot afford to lose until the receiver confirms the archive opens, because a typo in the password at packing time locks the files just as firmly as a forgotten one.

Start with the name. Name the archive after what is inside and when, for example a project and a date, instead of leaving it as "archive". The receiver sees that name long before they open it, and it keeps several versions apart in a downloads folder.

Folder layout decides what the receiver sees after extracting. Keep folders rebuilds the same tree you dropped, which matters for websites, code projects and anything that links files by path. All files at top level drops the folders and puts every file side by side, which is handier for a set of photos or scans that came from different folders.

Split into parts of cuts the archive into pieces of the size you set, in megabytes. Leave it at 0 for a single file. Split archives are useful for upload forms and mail servers with a fixed limit, but every part has to arrive before anything can be opened, so use the option only when a single file will not get through.

File names with accents, Cyrillic, Chinese or other non-English characters need one more thought. The ZIP format originally assumed the old IBM code page 437 for names, and APPNOTE later added a flag that marks names as UTF-8. Current systems read that flag, but very old extractors ignore it and show such names as strings of odd symbols. If the receiver reports garbled names, the files themselves are fine; opening the archive in a current app, or sending it as 7Z, which always stores Unicode names, fixes the display. To get files back out of an archive somebody sent you, use Unzip Online, and for other file jobs see more tools.