??? 06/13/06 11:59 Read: times |
#118217 - Fragmentation problem... Responding to: ???'s previous message |
Thank you for the reply,
Circular buffer would be cool, but as I mentioned my problem is with the fragmentation. Imagine this situation (the numbers are the packages sizes of bytes in the cyclic buffer g-get pointer, p-put pointer, f-free place): Now the buffer is full: 5 9 4 5 ^ g ^ p there are 2 timeouts of packages and I clear the first and the third package with size 4 and 5 bytes. So I get 9 bytes free. The next telegram that arrives has 9 bytes, I have 9 bytes free but I am still not able to save it in the buffer. F 9 F 5 ^ _ ^ g _ p 2. Each package if contains valid start and end characters, then do not store without a valid start and on the valid start... I didn't really underestand your second suggestion according start, end characters. Each package contains autonomous information for a specific sender. The package contains all the info the receiver needs. I can't have a start or end package, because it's possible that I receive only 1 package all the other are lost. But I also can't start processing the package until I am not sure that all the repeated package were received (upon timeout), because I would think I received 2x the same package (during processing). greetings Attila |
Topic | Author | Date |
Buffer management optimalization | 01/01/70 00:00 | |
Simple things first ... | 01/01/70 00:00 | |
You can use circular buffer | 01/01/70 00:00 | |
Fragmentation problem... | 01/01/70 00:00 | |
all methods have some problems | 01/01/70 00:00 | |
Start of package or End of package | 01/01/70 00:00 | |
Packages explanation | 01/01/70 00:00 | |
Individual buffers | 01/01/70 00:00 | |
Good Idea | 01/01/70 00:00 | |
New packages | 01/01/70 00:00 | |
hash table unefective | 01/01/70 00:00 | |
why keep that many | 01/01/70 00:00 | |
Reason of 5 or more buffer | 01/01/70 00:00 | |
Protocol specifics | 01/01/70 00:00 | |
One way protocol | 01/01/70 00:00 | |
One way protocol | 01/01/70 00:00 | |
never | 01/01/70 00:00 | |
Definition of need | 01/01/70 00:00 | |
Grant, I agree with what you post re thi | 01/01/70 00:00 | |
Simply reason why one way transmission | 01/01/70 00:00 | |
then why not just do it the easy way![]() | 01/01/70 00:00 |