Direct messages (personal messages sent on Twitter) should never post to your public timeline. This usually happens because of user error – meaning you posted the message like a public Tweet and our system sent it out as a Tweet.
Reasons this could have happened:
If you received a DM via text message, and replied without including "d username" at the beginning of your message, it is recognized as a public Tweet, not a direct message
If you typed your message into the Compose new Tweet box on Twitter but did not include "d username" at the beginning of the message, it will post as a public Tweet or @reply
Occasionally something will go wrong with a third party application, or a carrier will break a direct message into several parts so the last part is recognized as a public Tweet. (Those sending direct messages via text message/SMS should take care when using characters or letters from alphabets other than English, as these characters count as more than one English character.)
If you're not using a phone to post direct messages and they're still posting to the public timeline, it may an error with a third party application. Change your password immediately, and contact Twitter Support to alert folks to a possible bug in the app you are using.
Why are text messages in non-English alphabets broken into multiple parts?
If you're sending text messages in a language that uses non-English characters or accented letters, your messages obey a different character count
In addition to being measured by characters, a text message can also be measured in bytes
A standard text message is limited to 140 bytes. For the English alphabet, that equals 140 characters. For non-English alphabets it's a little different, as each character can require up to 4 bytes.
Example: " 我有一个没没 " has six characters, but consumes 18 bytes.
Example: "Voile est l'art du contrôle d'un bateau à voile." contains 48 characters but is actually 50 bytes.
Measurement by byte also occurs when typing updates into the web box. If you're typing in Chinese, the character count will measure the number of unicode bytes each character takes up to make sure that it's still under the equivalent of 140 English characters.
