A few days ago I saw someone by the name of p4bl0 mention that the Base64 encoding has a fixpoint. That is, if you repeatedly Base64-encode any string, the result will tend towards the same unique (infinite) string of characters.
This caught my interest. I wrote this C++ program to compute that fixpoint up to a specified precision (in sextets). It does this by, instead of repeatedly Base64-encoding a string, simply keeping track of the delta between the 8-bit input and 6-bit output (prior to table lookup).
cls@chalk$ ./b64fix 171 | fold -w 76; echo
Vm0wd2QyUXlVWGxWV0d4V1YwZDRWMVl3WkRSV01WbDNXa1JTVjAxV2JETlhhMUpUVmpBeFYySkVU
bGhoTVVwVVZtcEJlRll5U2tWVWJHaG9UVlZ3VlZadGNFSmxSbGw1VTJ0V1ZXSkhhRzlVVmxaM1Zs
WmFkR05GU214U2JHdzFWVEowVjFaWFNraGhSemxWVm14YU0xWnNXbUZrUjA1R1UyMTRVMkpIZHpG
cls@chalk$ ./b64fix 171 | base64
Vm0wd2QyUXlVWGxWV0d4V1YwZDRWMVl3WkRSV01WbDNXa1JTVjAxV2JETlhhMUpUVmpBeFYySkVU
bGhoTVVwVVZtcEJlRll5U2tWVWJHaG9UVlZ3VlZadGNFSmxSbGw1VTJ0V1ZXSkhhRzlVVmxaM1Zs
WmFkR05GU214U2JHdzFWVEowVjFaWFNraGhSemxWVm14YU0xWnNXbUZrUjA1R1UyMTRVMkpIZHpG
V1ZFb3dWakZhV0ZOcmFHaFNlbXhXVm0xNFlVMHhXbk5YYlVaclVqQTFSMVV5TVRSVk1rcElaSHBH
cls@chalk$
The result is reasonably fast:
cls@chalk$ n=$((32*1024*1024)) # 32 MiB
cls@chalk$ time ./b64fix $n > b64fix.txt
real 0m0.888s
user 0m0.772s
sys 0m0.099s
cls@chalk$ cmp -n $((n*4/3)) b64fix.txt <(base64 -w 0 b64fix.txt)
cls@chalk$
For more information, see https://news.ycombinator.com/item?id=37987225