TrueBrain · February 25, 2020 14:40
diff --git a/bat-shit-crazy-bananas-idea.txt b/bat-shit-crazy-bananas-idea.txt
 github.com/OpenTTD/BaNaNaS
 	Repository full with meta data
 	First level folders are the type (AI, Base Graphics, ..)
 	In these folders are folders with the IDs (in hex) of the uploads.
 	In there are "branches"; currently unused, but in there is for now a single .ini file.
 	These ini files are like described in https://github.com/OpenTTD/musa/blob/master/example.ini
 	They describe the current (latest) upload
 	Possibly a few references extra to find the file back on the CDN, but those are minor details
 		For example: MD5 + ID of upload should be in there too
 		MD5s will be encrypted, to avoid people finding the old MD5s of an ID. ottd_content and frontend are the only ones that needs to be able to decrypt it.
 	Repository admins can make any change needed
 	Also contains a blacklist, to avoid uploading certain IDs and MD5s (encrypted too, to avoid making it very easy for people to avoid the blacklist)
 	A bot will, after any change, update some administrative files
 		Per user there will be a small list with all the .ini files he/she has access to
 		A simple file for ottd_content to parse (it needs to know things like "tags" etc)

 cdn.openttd.org/BaNaNaS
 	Storage full with the binaries
 	First level folder are the type
 	Second level is the ID (in hex)
 	In there are all the uploads, based on the MD5
 	No listing of files on the CDN

 bananas.openttd.org
 	Public listing of all binaries (bit like now, but without the "downloads" column)
 		More information available, everything in the .ini files
 		Can decrypt the MD5s to generate URLs
 		API to serve all the information
 		Javascript to load everything dynamic
 		So no static HTML pages per entry (decreases complexity)
 	Manager-only part, login via GitHub
 		Small API backend that gives you back the ini files you have access to
 		Full javascript, you can edit/create an .ini file
 		Upload binaries via tus/tusd, directly to S3 (based on ID + MD5)
 		On submit, DorpsGek will make a commit in github.com/OpenTTD/BaNaNaS to update the latest state

 ottd_content
 	The server OpenTTD clients connect to (custom protocol)
 	Fetches a state-file from github.com/OpenTTD/BaNaNaS, and decrypts the MD5s in there
 	Returns "content_id" to refer to objects to download
 	Clients will contact balancer for downloads
 	ottd_content contains fallback method if HTTP doesn't work, to serve files over this custom protocol

 balancer
 	The server OpenTTD clients connec to in order to fetch files via HTTP
 	Client makes a POST with all the ids it wants to download, and expects a list of http links in return
 	Fetches a state-file from github.com/OpenTTD/BaNaNaS, and decrypts the MD5s in there

 musa
 	A CLI upload tool for BaNaNaS
 	Currently needs musad to run on the server in order to work
 	New version should simply wrap around the API
 	GitHub authentication might be the most trickest part
	github.com/OpenTTD/BaNaNaS
	Repository full with meta data
	First level folders are the type (AI, Base Graphics, ..)
	In these folders are folders with the IDs (in hex) of the uploads.
	In there are "branches"; currently unused, but in there is for now a single .ini file.
	These ini files are like described in https://github.com/OpenTTD/musa/blob/master/example.ini
	They describe the current (latest) upload
	Possibly a few references extra to find the file back on the CDN, but those are minor details
	For example: MD5 + ID of upload should be in there too
	MD5s will be encrypted, to avoid people finding the old MD5s of an ID. ottd_content and frontend are the only ones that needs to be able to decrypt it.
	Repository admins can make any change needed
	Also contains a blacklist, to avoid uploading certain IDs and MD5s (encrypted too, to avoid making it very easy for people to avoid the blacklist)
	A bot will, after any change, update some administrative files
	Per user there will be a small list with all the .ini files he/she has access to
	A simple file for ottd_content to parse (it needs to know things like "tags" etc)

	cdn.openttd.org/BaNaNaS
	Storage full with the binaries
	First level folder are the type
	Second level is the ID (in hex)
	In there are all the uploads, based on the MD5
	No listing of files on the CDN

	bananas.openttd.org
	Public listing of all binaries (bit like now, but without the "downloads" column)
	More information available, everything in the .ini files
	Can decrypt the MD5s to generate URLs
	API to serve all the information
	Javascript to load everything dynamic
	So no static HTML pages per entry (decreases complexity)
	Manager-only part, login via GitHub
	Small API backend that gives you back the ini files you have access to
	Full javascript, you can edit/create an .ini file
	Upload binaries via tus/tusd, directly to S3 (based on ID + MD5)
	On submit, DorpsGek will make a commit in github.com/OpenTTD/BaNaNaS to update the latest state

	ottd_content
	The server OpenTTD clients connect to (custom protocol)
	Fetches a state-file from github.com/OpenTTD/BaNaNaS, and decrypts the MD5s in there
	Returns "content_id" to refer to objects to download
	Clients will contact balancer for downloads
	ottd_content contains fallback method if HTTP doesn't work, to serve files over this custom protocol

	balancer
	The server OpenTTD clients connec to in order to fetch files via HTTP
	Client makes a POST with all the ids it wants to download, and expects a list of http links in return
	Fetches a state-file from github.com/OpenTTD/BaNaNaS, and decrypts the MD5s in there

	musa
	A CLI upload tool for BaNaNaS
	Currently needs musad to run on the server in order to work
	New version should simply wrap around the API
	GitHub authentication might be the most trickest part