Where base64 adds approximately 1/3, base85 only adds about
1/4. Of course there's a tradeoff. The Base85 alphabet includes
characters that might not be as friendly as the base64 alphabet. While it's
still only printable characters, the Ascii85 specification contains
") which needs escaping in many programming languages, and
the ZeroMQ specification contains
> which need
escaping in most (all?) SGML languages.
IPv6 encoding should only be used for encoding IPv6 addresses. When using IPv6, input for encoding must always be 16 bytes, and input for decoding must always be 20 bytes.
ZeroMQ's version (
z85) require according to the specification)
string input to be divisible by 5, and binary input to be divisible by 4.
Supported encoding specifications
npm install base85
var base85 = require('base85'); var z85 = base85.encode('Hello, world!!!!'); console.log(z85); // nm=QNz.92Pz/PV8aT50L
var base85 = require('base85'); var decoded = base85.decode('vqG:5Cw?IqayPd#az#9uAbn%daz>L5wPF#evpK6}vix96y?$k6z*rGH'); console.log(decoded.toString('utf8')); // all work and no play makes jack a dull boy!!
var base85 = require('base85'); var ipv6 = base85.encode('2001:db8:100:f101::1', 'ipv6'); console.log(ipv6); // 9R}vSQZ1W=8fRv3*HAqn var decoded = base85.decode('9R}vSQZ1W=8fRv3*HAqn', 'ipv6'); console.log(decoded); // 2001:db8:100:f101::1
Encodes the specified data. If encoding is
ascii85, the encoded data will be prepended with
<~and appended with
The data to encode, may be a
Stringor a Buffer.
Which specification to use when encoding
data. Valid values are:
ipv6. Default is
datais a buffer, it is expected to be the binary representation of an IPv6 address (16 bytes). It cannot be a textual representation. If it is a string, it can be on any valid IPv6 form (e.g.
1080:0:0:0:8:800:200c:417a, parsing is done using ip-address).
Stringwith the encoded data.
Decodes the specified data. If encoding is
ascii85, the data is expected to start with
<~and and end with
~>. No checks are actually made for this, but output will be unexpected if this is not the case. If encoding is
ipv6, the length of data must be exactly 20 bytes.
ipv6encoding cannot be used with arbitrary data.
A buffer is always returned as data may not be representable in a string. If you know it is, you can easily convert it to a string using the Buffer.toString() utility.
The data to decode. May be a
Stringor a Buffer. If
ascii85, it is expected to be enclosed in
datais encoded with. Valid values are:
ipv6. Default is
ZeroMQ appears to be a better specification for most applications. It doesn't
include quotes in its alphabet which makes it useful in many quoted languages
Neither does it add the 4 extra enclosing bytes Ascii85 does. There may,
however, be some problems using it in SGML and its derivatives since
< and greater-than
> are part of the alphabet. But
then again, Ascii85 has that as well.
Ascii85 appears to be the most used of the base85 specifications however. As for why completely eludes me. This may very well be the only reason to pick Ascii85.
If you control both decoding and encoding side, use ZeroMQ.
If you need interoperability with Ascii85, use that.
As IPv6 encoding only supports exactly 128 bits (16 bytes), this is not very useful for arbitrary data. Only use IPv6 if you're actually encoding IPv6 addresses.
Doesn't support the z-abbreviation for Ascii85. This means that data encoded with this
support will cause the library to return false. An all-zero input buffer will be encoded
<~!!!!!~>, rather than
IPv6 encoding specification (RFC1924) requires 128-bit arithmetic, which is rather problematic. I'm thrilled to see that the author of the RFC took this in consideration, specifically - quote from the RFC: "This is not considered a serious drawback in the representation, but a flaw of the processor designs." Silly processor designers. Currently, this is implemented using an arbitrary precision algorithm, it's slow but it does the job. Now let's poke those processor designers for 128-bit processors.