Background and motivation

Today, PemEncoding works on inputs of char, either arrays or spans, which implies UTF-16 encoded data. There has been the occasional need to work with UTF-8 encoded PEM structures, and will be needed in #91763.

This proposal is to extend PemEncoding to work with UTF-8 inputs and outputs.

Out of scope: new overloads of the many PEM APIs that have already been added throughout various cryptographic primitives will not gain UTF-8 overloads. Just the PemEncoding class.

API Proposal

namespace System.Security.Cryptography;

public static partial class PemEncoding {
+   public static PemFields FindUtf8(ReadOnlySpan<byte> pemData);
+   public static bool TryFindUtf8(ReadOnlySpan<byte> pemData, out PemFields fields);

+   public static bool TryWriteUtf8(ReadOnlySpan<byte> label, ReadOnlySpan<byte> data, Span<byte> destination, out int bytesWritten);
+   public static byte[] WriteUtf8(ReadOnlySpan<byte> label, System.ReadOnlySpan<byte> data);
}

Some remarks.

1. PemFields is entirely based on Range, which worked out well in this case so we don't need a new return type.
2. label is a ReadOnlySpan<byte>. It is expected that literals are almost always going to be used here though, like "CERTIFICATE"u8. We could continue to accept a char here instead, but would force use to do a UTF-16 to UTF-8 conversion. Since the compiler can do the conversion for us in the case of literals, it made sense to accept a UTF-8 ROS<byte>.

API Usage

Same as the existing UTF-16 inputs, except for UTF-8.

Alternative Designs

No response

Risks

No response