1

Topic: Fast ID for objects of the network protocol

For example there is at us a protocol where a command it ObjectcId. Property = Value it is interesting to me as it is possible to organize the fastest passage from object to its pointer, both on the server and on the client 1. To generate one GUID but then it is necessary to search for everyone  in the table at each command 2. It is stupid to exchange addresses in storage - to service to tell where the object lies on the client and on the contrary then it is possible to result simply  in the pointer that , but the request-answer here is necessary, and these are brakes global uniqueness will be provided with uniqueness of pair  process (or , or sessions) and addresses in storage as  - the client creates object and sends its address, and the server places it in the table, and on the contrary 3. To select accessible one  as HANDLE in Windows, here access fast - it is simple on an index in an array but if the object forms by the client that   to provide  with  the table, or again to wait    better to eat, I personally like idea of the address then not  on each command  to do 4. Still invented :  creates a pool of objects in advance and delivers clients of a pack accessible  when the client settled a pool - he asks a new pack of addresses this decision  zarpos-answer brakes through preliminary  at once heaps of objects well and here we will have a storage over-expenditure on the server

2

Re: Fast ID for objects of the network protocol

Hello, Barbar1an, you wrote: B> it is interesting to me as it is possible to organize the fastest passage from object to its pointer, both on the server and on the client you catch fleas. You are assured what exactly this bottleneck of your server? B> 1. To generate one GUID but then it is necessary to search for everyone  in the table at each command cache? B> 2. It is stupid to exchange addresses in storage - to service to tell where the object lies on the client and on the contrary then it is possible to result simply  in the pointer that , but the request-answer here is necessary, and these are brakes B> global uniqueness will be provided with uniqueness of pair  process (or , or sessions) and addresses in storage And still it is necessary to check not  sent as the address. B> as  - the client creates object and sends its address, and the server places it in the table, and on the contrary Kopromis simply index in an array it handle.

3

Re: Fast ID for objects of the network protocol

Hello, Barbar1an, you wrote: B> for example there is at us a protocol where a command it ObjectcId. Property = Value B> it is interesting to me as it is possible to organize the fastest passage from object to its pointer, both on the server and on the client If you have a server and the client divided spatially any of your variants will occupy 0.00. 01 % from operation Therefore the correct variant - 1 (to use GUID), remaining - delirium

4

Re: Fast ID for objects of the network protocol

B> 2. It is stupid to exchange addresses in storage - to service to tell where the object lies on the client and on the contrary then it is possible to result simply  in the pointer to Hackers such protocol definitely it is pleasant