Currently I am in the planning stages for when I plan to run a game of D&D 4E, but this question its self is more system-agnostic.
I want to design a written language that will come up in some of the puzzles I present to the players. Some of these puzzles may be alongside other similar puzzles, while some would be the main focus. The main thing is I want to use a persistent language, and not just random symbols each time.
Ignoring coming up with such a puzzle language(a task for another time), I am in need of ideas on how to present such a puzzle, but not make it break future puzzles involving the same language(aka a meta-gamer bringing his own decoded language reference which is more than the characters have gathered in game). While I don't want to limit creative thinking on these puzzles, I just need ideas to keep them from becoming stale.
–UPDATE–
I feel the wrong question is being answered. How to design a secret language is not what I need help with, but that is what is being answered.
What I need ideas for is implementing a language as part of various puzzles, where success or failure at a previous puzzle could affect the outcomes of future puzzles involving the language. I don't want to make it to easy to early, but at the same time I don't want making a single failed puzzle fail the whole lineup.
Best Answer
This is a question of character knowledge versus player knowledge.
Cryptanalysis is a fun, if little used, human skill. Some players will delight in solving cryptographic puzzles, and 4e certainly doesn't prohibit the gamist approach of "challenge the player."
On the other hand, cryptanalysis is highly binary. People either really like it, or basically can't stand it, especially if you start branching out into some of the more interesting, but trivially solved cyphers (playfair, Vigenère).
I, unfortunately, had to learn this fact the hard way. Players, as a rule, will take one of two approaches to puzzles. The first approach I learned when I gave a cryptogram in a sci-fi game: "I hire an expert from town to solve it for me." is the general pattern here. To make this approach interesting, the cryptanalysis must have extrinsic complications such that one can't submit it to an analyst (magical or otherwise) and wait it out. Beyond that, having a player go "and I solve it. what does it say?" is absolutely valid, and should require either a pure int check, or a skill check of an appropriate skill the player is bringing to bear. I would personally recommend looking at the serious skills series to see how to make this approach more interesting, as players can add depth to their character by describing how a particular facet of knowledge was brought into play. Still, if you're not expecting it, it comes as a horrible horrible surprise. Some people just don't find puzzles fun.
For the players who find puzzles fun, especially for a 4e game, take a page from Portal.
Teach players one component at a time of the puzzle. It'll seem way too easy to you, but you know the context. For a crypto puzzle, start by teaching the alphabet. (Maybe touch the "true characters" not to get shocked) In subsequent challenges, teach the order of the alphabet. Again, always give an example of what you're looking for before you ask for what you're looking for. (Or, provide a way for rapid abductive reasoning without penalty.)
This way, after quite a few sessions, your players will be able to manipulate the language as a tool to solve problems, and may occasionally be assisted by their characters-as-characters.
An interesting touch would be to make the language three dimensional, and have the puzzles follow the rules of the game Zendo. Then the rough series of questions can be "complete this sentence." or "find the incorrect thing." Both are completely appropriate to solving a crypto-language, and both can engage players, instead of abstracting the puzzle to their characters.