We don't need any new rules: the help center already endorses fixing spelling mistakes.
Any such changes should be individually justified and should not counteract the author's reasonable style choices. The examples listed do not make a good set of rules for slavish adherence.
- Different spellings of ElGamal
The cryptographer (and now CTO of Salesforce) spells his name Taher Elgamal in English. The original Arabic is طاهر الجمل, which does not have case distinctions. One could reasonably read الجمل with a definite article and transliterate it as ‘al-Jamal’ just as well as transliteratingone could reasonably transliterate it intowith the anglophone typographical conventions for names as ‘Elgamal’. ‘ElGamal’ also appears in print, in publications before he decided that the intermediate uppercase letter was too much trouble to deal with.
While ‘al-Jamal’ may not be recognizable in cryptography, ‘Elgamal’ and ‘ElGamal’ are both justifiable and we should not edit one to the other.
- Spelling of Vigenère with 'e' instead of 'é';
Blaise de Vigenère, the diplomat and cryptographer of the XVIe century to whom the Vigenère cipher is misattributed, had his name typeset in publications as Blaiſe de Vigenere (example; note the use of accents elsewhere in the same publication, suggesting a conscious decision).
While there is nothing wrong with writing Vigenère to reflect the orthography for modern Parisian pronunciation, I see no reason impose upon ourselves a rule to bow before the authoritarian prescriptions of l'Académie française; the Vigenère cipher is just as recognizable as the Vigenere cipher.
- Aes-128 where the acronym is not in caps;
Aes-128 is an abomination.
- AES128 or SHA256 where the dash is missing;
The dash is appropriate in some contexts and inappropriate in others. In isolation it is a little clearer on the eyes to talk about SHA-256 and SHA3-256 than than to talk about SHA256 and SHA3256; but while SHA3-256 is nicer than SHA-3-256, SHA-3 is nicer than SHA3; and it is clearer to say HMAC-SHA256 than to say HMAC-SHA-256—yet with the possible exception of SHA3256 (which sometimes appears in programming interfaces and is not much more wrong than Poly1305 to refer to $2^{130} - 5$ or Curve25519 to refer to $2^{255} - 19$) none of these is wrong.
A foolish consistency is the hobgoblin of professional minds.