This is the decode function if anyone is interested:

decoded_reference()
{
    local yr_msd=${1:0:1}
    local yr_lsd=${1:1:1}
    local seq_enc_msd=${1:3:1}
    local seq_enc_lsd=${1:4:1}
    local seq_msd=${lookup_table_reverse[$seq_enc_msd]}
    local seq_lsd=${lookup_table_reverse[$seq_enc_lsd]}
    local seq_msd_index=$(typeset -p symbolset | grep -oP '[0-9]+(?=]="'"$seq_msd"'")')
    local seq_lsd_index=$(typeset -p symbolset | grep -oP '[0-9]+(?=]="'"$seq_lsd"'")')
    local seq=$((seq_msd_index * ln_symbolset + seq_lsd_index))
    local yr_msd_index=$(typeset -p symbolset | grep -oP '[0-9]+(?=]="'"$yr_msd"'")')
    local yr_lsd_index=$(typeset -p symbolset | grep -oP '[0-9]+(?=]="'"$yr_lsd"'")')
    local yr=$((ln_symbolset * ln_symbolset * 2 + yr_msd_index * ln_symbolset + yr_lsd_index)); # warning: the “2” is a dangerous hard-coding! Hopefully that bug manifests after I am dead

    printf '%s\n' "${yr}-$seq"
};#decoded_reference

I probably need a perfect hash function. This code seems to do the job:

encoded_reference()
{
    local -r yr=$1
    local -r seqno=$2
    
    local -ar symbolset=(a b c d e f g h   j k   m n   p q r s t u v w x y z     2 3 4 5 6 7 8 9)
    local -a seedset=("${symbolset[@]}")
    local -r ln_symbolset=${#symbolset[@]}; # 31
    local ln_seedset=${#seedset[@]}
    local -A lookup_table=()

    for sym in "${symbolset[@]}"
    do
        pos=$((50 % ln_seedset)); # 50 is just an arbitrary static number
        lookup_table+=(["$sym"]=${seedset["$pos"]})
        seedset=(${seedset[@]/${seedset[$pos]}}); # remove used elements from the seedset
        ln_seedset=${#seedset[@]}
    done
    
    local yr_enc=${symbolset[$(((yr / ln_symbolset) % ln_symbolset))]}${symbolset[$(($yr % ln_symbolset))]}
    local most_sig_fig=$((seqno / ln_symbolset))
    local least_sig_fig=$((seqno % ln_symbolset))
    
    # caution: if the seqno exceeds ln_symbolset², this calculation is out of range
    local seq_enc=${lookup_table[${symbolset[$most_sig_fig]}]}${lookup_table[${symbolset[$least_sig_fig]}]}
    
    printf '%s\n' "answer → ${yr_enc}-$seq_enc"
};#encoded_reference

for yr in 2024 2025 2026
do
    for seqno in {1..20}
    do
        encoded_reference "$yr" "$seqno"
    done
done

This is close to ideal, but I just thought of another problem: what if a year-seq pair were to derive an encoded number like “fy-ou” or “us-uk” or “sh-it”? A bias that nearly ensures a digit is used would help avoid generating offending words. But I guess I’m getting well into over-engineering territory.

That is certainly a winner from the standpoint of code simplicity. And it’s trivially reversible. But I’m also prioritizing simplicity for human recipients above code simplicity. Base64 output is case sensitive and someone writing back and referencing a ref number would not necessarily preserve case. It’s also intolerant of human errors like confusing a “1” for a “l”.

(edit) I think base32 would avoid the case sensitivity problem. So here’s a sample:

for seq in {1..60}; do printf '%s → ' 2024-"$seq"; printf '%s\n' 2024-"$seq" | base32 | awk '{print tolower($1)}' | sed 's/=//g'; done

2024-1 → giydenbngefa
2024-2 → giydenbngifa
2024-3 → giydenbngmfa
2024-4 → giydenbngqfa
2024-5 → giydenbngufa
2024-6 → giydenbngyfa
2024-7 → giydenbng4fa
2024-8 → giydenbnhafa
2024-9 → giydenbnhefa
2024-10 → giydenbngeyau
2024-11 → giydenbngeyqu
2024-12 → giydenbngezau
2024-13 → giydenbngezqu
2024-14 → giydenbnge2au
2024-15 → giydenbnge2qu
2024-16 → giydenbnge3au
2024-17 → giydenbnge3qu
2024-18 → giydenbnge4au
2024-19 → giydenbnge4qu
2024-20 → giydenbngiyau
2024-21 → giydenbngiyqu
2024-22 → giydenbngizau
2024-23 → giydenbngizqu
2024-24 → giydenbngi2au
2024-25 → giydenbngi2qu

The “js” example is just to encode the year which is a prefix to the encoded sequence number. So if 2024 gives “js”, then ref numbers would look like this:
js-aa
js-ab
js-ac
…etc.

And I do not reset the counter at the beginning of the year. So 2025 would be like:

jt-ad
jt-ae
jt-af
…etc.

(update)

Rereading, maybe I misunderstood - would the full string include the date? so 2024-js as a complete example?

Yes, but note that “js” in my example was for an encoding of the year, which helps shrink the reference number and mask the fact that the 2nd token is a sequence.

Your client would make a difference. What you are probably seeing is the mirrored version of !tex@lemmy.sdfeu.org on lemm.ee. You cannot possibly be interacting with a non-existent community. If I post to https://linkage.ds8.zone/c/tex@lemmy.sdfeu.org, then I don’t suppose you would see it on https://lemm.ee/c/tex@lemmy.sdfeu.org.

(edit) just saw your test msg. Well, that’s interesting. Even though !tex@lemmy.sdfeu.org no longer exists, it seems the mirrored versions of it can still collaborate. I’m not sure how that works.

I’m not sure what that is. vger.social and voyager.lemmy.ml don’t seem to have anything relevant. But I found !tex_typesetting@lemmy.sdf.org.