Return-Path: <cyrus@holtmann.org>
MIME-Version: 1.0
In-Reply-To: <20120928104318.GC25415@x220>
References: <1348584763-22824-1-git-send-email-andrzej.kaczmarek@tieto.com>
	<1348584763-22824-8-git-send-email-andrzej.kaczmarek@tieto.com>
	<20120928104318.GC25415@x220>
Date: Fri, 28 Sep 2012 13:09:54 +0200
Message-ID: <CACLukzLZGEbA-3pSqbutRrjb5WXJo7az+DydQ1Vph5bAnD0kdQ@mail.gmail.com>
Subject: Re: [PATCH 07/14] thermometer: Reformat MeasurementReceived description
From: Santiago Carot <sancane@gmail.com>
To: Andrzej Kaczmarek <andrzej.kaczmarek@tieto.com>,
	linux-bluetooth@vger.kernel.org
Content-Type: text/plain; charset=ISO-8859-1
Sender: linux-bluetooth-owner@vger.kernel.org
List-ID: <linux-bluetooth.vger.kernel.org>

Hi johan,

2012/9/28 Johan Hedberg <johan.hedberg@gmail.com>:
> Hi Andrzej,
>
> On Tue, Sep 25, 2012, Andrzej Kaczmarek wrote:
>> +                     Measurement:
>> +
>> +                             int16 Exponent:
>> +                             int32 Mantissa:
>> +
>> +                                     Measurement value is calculated as
>> +                                     (Mantissa) x (10^Exponent)
>> +
>> +                                     In case of invalid or missing data,
>> +                                     Exponent is set to 0 and Mantissa is
>> +                                     set to 2^23-1 (0x7FFFFF hexadecimal).
>
> Why aren't we using the D-Bus DOUBLE type for this? For invalid/missing
> data just leave out these values from the dict. No need to try to
> brute-force a 1:1 mapping from the protocol to D-Bus.
>

We don't use double type here because these information is valuable
for upper profiles implementing a IEE11073-20601 protocol layer. This
fact is scarcely commented in HTP and better explained in former
specification where there is a need to pack both values (mantissa and
exponent) in special kind of APDUs. Of course we could provide the
double value here, but there will remain the need to know when an
special case occurs: NRes, NaN,INFINITY, -INFINITY.

Furthermore, providing a double value here will force to upper
profiles to do extra stuff to calculate the mantissa and the exponent
values in order to create APDUs, beside they will have to deal with
impreccisions derivated to comparate float point numbers and so it
might make difficult to check when a special case has happened.

these are, as far as I remember, the reasons which we did it in this
way. We were already talking about this fact in the mailing list in
the past but I don't find the thread right now. sorry