protocol execution: fix party index

[lenerd]

Introduced in 79610b4, this bug
resulted in miscalculated party indices starting from -1:

MATRIX/Execution/fabfile.py

Line 111 in 79610b4

run('. ./%s %s %s' % (executable_name[idx], party_id - 1, values_str))

[liorko87]

It's not mistake. Take a look at this scenario:
Let's assume that we have 3 parties that act as clients and one server that coordinate between them, total of 4 different machines.
The server needs to be on air when the parties tries to connect., so he will be the first party, i.e. party_id= 0. The clients id are running from 0 to 2.
because the server defined as party_id= 0 and the clients are also need to start from 0 the line is run('. ./%s %s %s' % (executable_name[idx], party_id - 1, values_str)) and not run('. ./%s %s %s' % (executable_name[idx], party_id , values_str))

I will add some description to the fabfile at the next commit

[lenerd]

Hi Lior,

the line I referenced above handled the case without a coordinating party.

However, you have fixed the off-by-one in ef6ef8a here:

• ef6ef8a#diff-c437b24436406fb96d62b02c66bee442L115
• ef6ef8a#diff-c437b24436406fb96d62b02c66bee442R123

Thanks!

[lenerd]

With the changes from ef6ef8a, the parties.conf file is no longer uploaded to the servers if the (external) protocol is run without coordinating parties. See

• ef6ef8a#diff-c437b24436406fb96d62b02c66bee442L113

and maybe

• ef6ef8a#diff-c437b24436406fb96d62b02c66bee442L109

Edit: See #29.