5 ms·

Can someone give a plausible explanation why base64 is used instead of base85? Both are transcoded in linear time (Even base91 is possible in linear time but po

by aydyn 3y ago

Can someone give a plausible explanation why base64 is used instead of base85? Both are transcoded in linear time (Even base91 is possible in linear time but potentially still slower due to more difficult SIMD). People have been making the argument for over a decade that base64 is incumbent and so people stick with it due to interoperability. But base85 represents a 20% compression gain for basically free from a computational perspective. Isn't that worth switching over as a widely used standard?

chubot 3y ago
Was it free when base64 was invented? CPUs were slower relative to memory in those days, so it may have made a difference. Many more programs were CPU bound, and people wrote assembly more often Also base64 is just easier to code
kps 3y ago
One of the reasons for the shift from uuencoding to base64 is that the latter uses almost exclusively alphanumeric characters and consequently is more robust against cross-platform or ‘smart’ transformations or quoting or escaping issues.
barryrandall 3y ago
20% better isn't enough to motivate every single producer and consumer to incur the switching costs or taking on an IPv6-scale migration process.
layer8 3y ago
Base85 uses ASCII characters 33 (“!”) through 117 (“u”), which includes a number of inconvenient characters, like backslash and single and double quotes. The reason to use an encoding at all instead of just raw binary is not just to avoid nonprintable characters, but also to avoid certain printable characters with meta semantics, like string delimiters and escape characters. An alternative would be the encoding specified by RFC 1924, which uses a different, noncontiguous set of characters. It still has the drawback that dividing by 85 is a bit slower than dividing by 64 (which is just a bit shift). Last but not least, Base64 has the benefit of being easily recognizable by a human. Due to its relatively restricted character set, it doesn’t look like just line noise, but also doesn’t look like more intelligible syntax, or like hex, etc. It sits in a middle sweet spot.

by aydyn 3y ago

chubot 3y ago

Was it free when base64 was invented? CPUs were slower relative to memory in those days, so it may have made a difference. Many more programs were CPU bound, and people wrote assembly more often Also base64 is just easier to code

kps 3y ago

One of the reasons for the shift from uuencoding to base64 is that the latter uses almost exclusively alphanumeric characters and consequently is more robust against cross-platform or ‘smart’ transformations or quoting or escaping issues.

barryrandall 3y ago

20% better isn't enough to motivate every single producer and consumer to incur the switching costs or taking on an IPv6-scale migration process.

layer8 3y ago

Base85 uses ASCII characters 33 (“!”) through 117 (“u”), which includes a number of inconvenient characters, like backslash and single and double quotes. The reason to use an encoding at all instead of just raw binary is not just to avoid nonprintable characters, but also to avoid certain printable characters with meta semantics, like string delimiters and escape characters. An alternative would be the encoding specified by RFC 1924, which uses a different, noncontiguous set of characters. It still has the drawback that dividing by 85 is a bit slower than dividing by 64 (which is just a bit shift). Last but not least, Base64 has the benefit of being easily recognizable by a human. Due to its relatively restricted character set, it doesn’t look like just line noise, but also doesn’t look like more intelligible syntax, or like hex, etc. It sits in a middle sweet spot.