I don't agree with that. PB is practically the same as DER. All the attack surface area lies in the codec, specifically in the decoder.

▲

mananaysiempre 3 months ago | parent [-]

There are two notions of surface area here: that exposed to the external input, which must be secured, and that exposed to the programmer, which must be understood. As far as the latter is concerned, one can’t really disassociate the encoding of DER from the, well, notation of ASN.1, which, while definitely not as foreign as it may first appear, is still very rich compared to the one Protobufs use. (I do think a good tutorial and a cheat-sheet comparison to more widely used IDLs would help—for certain, obscure dusty corners and jargon-laden specs have never stopped anyone from writing the COM dialect of DCE IDL.)

Even if we restrict ourselves to the former notion, the simple first stage of parsing that handles DER proper is not the only one we have to contend with: we also have to translate things like strings, dates, and times to ones the embedding environment commonly uses. Like, I’m the kind of weird pervert that would find it fun to implement transcoding between T.61 and Unicode faithfully, but has anyone ever actually put T.61 in an ASN.1 T61String? As far as I know, not as far as PKIX is concerned—seemingly every T61String in a certificate just has ISO 8859-1 or *shudder* even Windows-1252 inside it (and that’s part of the reason T61Strings are flat out prohibited in today’s Web PKI, but who can tell about private PKIs?). And I’ll have to answer questions like this about every one of a dozen obscure and/or antiquated data types that core ASN.1 has (EMBEDDED PDV anyone?..).

	▲	cryptonector 3 months ago \| parent [-]
		> There are two notions of surface area here: that exposed to the external input, which must be secured, and that exposed to the programmer, which must be understood. As far as the latter is concerned, one can’t really disassociate the encoding of DER from the, well, notation of ASN.1, [...] I disagree. I say that as a part-time maintainer of an open source ASN.1 stack that generates ergonomic C from ASN.1. > I do think a good tutorial and a cheat-sheet comparison [...] For ASN.1? There's a great deal of content out there, and several books. I'm not sure what more can be done. Tutorials? Look around this thread. People who can't be bothered with docs nowadays also can't be bothered with tutorials -- they just rely on LLMs. > Like, I’m the kind of weird pervert that would find it fun to implement transcoding between T.61 and Unicode faithfully, but has anyone ever actually put T.61 in an ASN.1 T61String? Me too, but as you note, no one really does that. My approach as to PKIX is to only-allow-ASCII for string types other than UTF8String. > And I’ll have to answer questions like this about every one of a dozen obscure and/or antiquated data types that core ASN.1 has (EMBEDDED PDV anyone?..). Now do C++! A "modern" subset of ASN.1 is not that much smaller than x.680 + all of x.681, x.682, and x.683.