This is the mail archive of the cygwin mailing list for the Cygwin project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: hacked package on server

I remember an extensive discussion about the purpose MD5 sums serve for 
the cygwin installer[1] some time ago.

My understanding (synopsis) of the thread (before I gave up reading it in 
disgust somewhere around Christopher Faylor's suggestion that the subject 
be dropped :} ) was that md5sum is only used to indicate that a particular 
package had been completely downloaded (someone correct me if I 
misunderstood, please).  If that is the case, then it seems to me to be a 
bit quick to declare a server or package compromised based on a mismatch 
of md5 sums.

I do think that instead of simply aborting the install with a message that 
the server was compromised (was it?  or is something else going on?), that 
a more useful option would be to allow the user to select a different 
mirror and continue the process.

William Sutton


On Mon, 16 Jul 2007, Louis Kruger wrote:

> >
> > On Mon, Jul 16, 2007 at 10:30:52AM -0500, Louis Kruger wrote:
> > > I also have a complaint:  the dialog that notifies the user of the failed 
> > > MD5 is not well designed.  The dialog asks "Do you want to skip the 
> > > package?" and has a yes and no button.  I read it quickly and pressed no 
> > > before thinking about it, the package went ahead and tried to install.  I 
> > > think there should be a little more effort to restrain the user from 
> > > performing a dangerous action such as installing a package with a wrong MD5.
> >
> > Good point.  The message should probably be
> >
> > Do you want to not skip the package (No/Yes)?
> >
> > cgf
> I realize you are joking, but the wording of the message is beside the 
> point.  For an ordinary end-user, installing a file with a wrong MD5 is 
> the wrong (and dangerous) thing to do in just about any case I can think 
> of.  Therefore it should not be equally easy to select either option.
> My opinion is that the setup program should abort immediately on 
> detecting a wrong MD5 with a message that the server may have been 
> compromised.  If there is a special case where someone may actually want 
> this, it should be something non-obvious, like a -allow-wrong-md5 flag 
> to the setup program.
> thanks,
> Louis
> --
> Unsubscribe info:
> Problem reports:
> Documentation:
> FAQ:         

Unsubscribe info:
Problem reports:

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]