From mboxrd@z Thu Jan 1 00:00:00 1970 From: Bennett Todd Subject: filenames that can be safely stolen (was Re: starting with ".." could break stuff) Date: Fri, 2 Apr 2004 22:02:50 +0000 Message-ID: <20040402220250.GB4668@rahul.net> References: <200404010546.i315kEvY019805@sirius.cs.pdx.edu> <20040402035617.GC2158@zero> <8765ciqr6v.fsf@uhoreg.ca> Mime-Version: 1.0 Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="rS8CxjVDS/+yyDmU" Return-path: list-help: list-unsubscribe: list-post: Errors-To: flx@namesys.com Content-Disposition: inline In-Reply-To: <8765ciqr6v.fsf@uhoreg.ca> List-Id: To: reiserfs-list@namesys.com --rS8CxjVDS/+yyDmU Content-Type: text/plain; charset=us-ascii Content-Disposition: inline Content-Transfer-Encoding: quoted-printable 2004-04-02T16:29:12 Hubert Chan: > Tom> i bet there's a few apps that, when trying to skip .., only check > Tom> the first two chars of the name. this may or may not be a problem. >=20 > Then those apps are broken. ..foo has always been a valid filename > (though highly uncommon). ..metas doesn't introduce anything new -- it > only makes ..files more common. Yup. For another example, AFS steals "@sys"; it's a special black-magic voodoo name that acts like a symlink to a machine-specific value, used to set up baroque symlink trees so one uniform network filesystem can support dispatching to appropriate versions of machine-specific files like executables. So another possibility worth considering might be @metas, with the prior art of AFS to cite to people who complain about theoretical problems:-). -Bennett --rS8CxjVDS/+yyDmU Content-Type: application/pgp-signature Content-Disposition: inline -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.0.7 (GNU/Linux) iD8DBQFAbeMJHZWg9mCTffwRApicAJ47at8dQhi44MS6cSKp8nbjnP8HEgCeMxRL 2A1DyCYU4KTDUcDW4VKZ8lo= =4J97 -----END PGP SIGNATURE----- --rS8CxjVDS/+yyDmU--