This seems like an obvious question but I haven't been able to find it, so here goes:
Enigma is based on an alphabet of A-Z (26 characters); obviously this directly affects message content (e.g. numbers would be specified using their names one, four, two, etc) and it also has a direct bearing on the number of possible permutations (crypto strength).
Would increasing the alphabet of the rotors (e.g. to A-Z, 0-9, and a selection of special characters) increase crypto strength? I know it would increase the number of possible permutations but I'm not sure if that would introduce some kind of weakness.
Update (additional context)
I'm building a "for fun" software implementation of Enigma, in C# - currently as a UWP application that will run on any Windows 10 device (including phones). It will include implementations of a few of the real-world historical Enigmas (I, M3, M4) as well as the ability to take "extended" Enigmas where you can create your own rotor sets with any number of rotors, including new wiring's and character sets (i.e. beyond A-Z).
Based on feedback provided adding additional characters could be dangerous.
I was thinking that for an 'extended' implementation you might want to include 0-9 and some special characters commonly used these days (such as : /). I guess one motivation for this was convenience. Of course given the flexibility of modern computing I guess I'd be better off allowing the user to input "convenience-plain-text" such as "HTTP : //WWW" transform it to "standard enigma A-Z plain-text" "HTTPCOLONBACKSLASHBACKSLASHWWW" then encipher it, and possibly perform the reverse at the other end. Maybe that's the basis of further questions another time.
Update - 28-July-2017
Based on this Q&A I can now actually measure the increase in key-size for enigma like systems; whilst expanding the supported character set does increase the key size, adding additional rotors increases it by significantly more (notwithstanding the other issues that affect the overall crypto-strength of enigma).