Proposal: Support for RFC4648 in PHP

Why do it in core?
Is userland base16/base32 performance unsatisfactory?

Even if the performance is satisfactory, having this in core can help
user-land code to get rid of their own implementations (which they
often do, to be URL-safe, for example).

I imagine native implementations can make use of CPU SIMD acceleration
as well, like we already do for the current base64 functions.

The initial message also mentions adding this functionality into a
separate namespace, so I'm quite positive with this proposal. Even
more so than the previous proposals that proposed to add new
parameters to the base64 functions.

On Thu, Jun 12, 2025 at 7:23 PM Hans Henrik Bergan divinity76@gmail.com
wrote:

On Thu, 12 Jun 2025 at 18:53, ignace nyamagana butera
nyamsprod@gmail.com wrote:

Greetings,

I have been playing around with an RFC proposal for some time and
recently after discussing it with Tim Düsterhus he has volunteered
to do the implementation so I would like to submit to you an RFC around
data encoding in PHP. The goal of the RFC is to fully implement
RFC4648 (base16 , base32, base64).

I know that PHP already has base64_encode and base64_decode and
bih2hex and hex2bin but those functions only provide a partial support
for RFC4648.
In my RFC proposal I would like to introduce, instead, a new namespace
Encoding that will host an all new and improved API which will cover the
complete RFC,
will be consistent and easily extendable for future data encoding
algorithms addition.

For reference, this is not a new topic, the issue with the current
implementation is well documented in the mailing list as well as a past
attempt to
add RFC4648 to the language.

• base64_encode without padding https://externals.io/message/122630
• base64 url format https://externals.io/message/119243
• [RFC] RFC4648 encoding https://externals.io/message/91858#91964

If people are interested I will proceed with a karma request and create
the draft RFC.

Best regards,
Ignace Nyamagana Butera

Why do it in core?
Is userland base16/base32 performance unsatisfactory?

Hi Hans,

If I take base32 as an example, if you look at packagist for userland
implementations packages, you will get different results with none telling
you which version (alphabet) is being used
and if the decoding is strict or not.. The same is true for base64, there
are different flavours of base64 with each their usefulness. The RFC will
address these issues and provide a consistent API to tackle them.

Best regards,
Ignace Nyamagana Butera