El mié, 04-02-2004 a las 04:02, Olaf Frączyk escribió: > So, maybe we should just wait? > Such databases will be really huge, as having millions of files is > nothing unusual. Just for mime types you will need about 150MB per 1 > million of files. (It is with assumption that per average file we use > 150 bytes). Next think if you have 50 users or more. I have 139427 files. I'd gladly pay the 15 MB penalty for having more functionality. Those who don't want to pay for it don't need to. > Next think of users with quotas. > > Personally I think that it is not worth to write per user solution. It > will be slower, it will take too much space on disk, it will be less > secure. With multi-user solution most of examining may be done at night, > when nobody is working. You're confusing the problems of: * indexing file metadata * indexing file data * managing file metadata The management of metadata we can leave to the end user, perhaps with some automation on the user apps' part (mp3 taggers, openoffice, all such apps should use the metadata api to enrich filesystem presence for all their files). The indexing of metadata and data is what you're referring to examination. The key difference between what you're proposing and what has already been done is that you're proposing to store the metadata *separate* from the files. This is a mistake, plain and simple. You'll introduce new vulnerability sources and lower the robustness of the system. > > Regards, > > Olaf -- Manuel Amador (Rudd-O) GPG key ID: 0xC1033CAD at keyserver.net
Attachment:
signature.asc
Description: Esta parte del mensaje =?ISO-8859-1?Q?est=E1?= firmada digitalmente