1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430
|
\section{Specification}
This case study describes the definitions and theorems needed to build a
complete sliding window protocol theory. The files needed to do this are in
the directory:
\[ \verb%hol/Training/studies/protocols%. \]
All the input commands which are given in boxes can be found in the file
\verb%SWnew.ml%. To work through this case study in HOL edit the file
\verb%SWnew.ml% while running HOL and read through this document, entering
commands as they appear in the text.
To build the sliding window protocol theory, first enter HOL and set up the
sliding window theory :
\setcounter{sessioncount}{1}
\begin{session}
\begin{verbatim}
loadf`startSW`;;
\end{verbatim}
\end{session}
This command will open a new theory and load its parent theories.
The parent theories for this sliding window protocol theory
are, \verb%myarith% which contains some necessary
arithmetic theorems, \verb%plusm_subm% which contains a theory of
modular arithmetic and \verb%hdi_tli% which contains definitions and
theorems for operations on sections of lists.
A summary of the theories \verb%hdi_tli% and \verb%plusm_subm% can be
found in Appendix A.
The structure of the HOL model to be specified is illustrated in
Figure~\ref{fig:holstruct}.
The boxes in the diagram represent HOL predicates and
the arrows represent the flow of information or data in the system.
\begin{figure}
\setlength{\unitlength}{1mm}
\small
\begin{picture}(150,70)
\put(15,0){
\begin{picture}(120,70)
\put(45,3){\framebox(30,7){{\sf CHANNEL} (ack)}}
\put(23,55){\framebox(14,7){{\sf ABORT}}}
\put(30,50){\line(0,1){5}}
\put(15,20){\framebox(30,30){}}
\put(19,43){\makebox(22,6){{\sf SENDER}}}
\put(19,35){\framebox(22,7){{\sf DataTrans}}}
\put(27,32){s}
\put(32,32){rem}
\put(30,30){\vector(0,1){5}}
\put(19,23){\framebox(22,7){{\sf AckRecv}}}
\put(75,20){\framebox(30,30){}}
\put(79,43){\makebox(22,6){{\sf RECEIVER}}}
\put(79,35){\framebox(22,7){{\sf DataRecv}}}
\put(92,32){r}
\put(90,35){\vector(0,-1){5}}
\put(79,23){\framebox(22,7){{\sf AckTrans}}}
\put(55,29){\framebox(10,6){{\sf INIT}}}
\put(45,55){\framebox(30,7){{\sf CHANNEL} (data)}}
\put(10,26){\vector(1,0){8}}
\put(6,28){\makebox(0,0){source}}
\put(6,24){\makebox(0,0){= rem 0}}
\put(102,38){\vector(1,0){8}}
\put(114,38){\makebox(0,0){sink}}
\put(41,42){\vector(2,3){8}}
\put(38,52){dataS}
\put(74,52){dataR}
\put(71,55){\vector(2,-3){8}}
\put(49,10){\vector(-2,3){8}}
\put(38,13){ackS}
\put(74,13){ackR}
\put(79,23){\vector(-2,-3){8}}
\put(65,29){\vector(4,-1){14}}
\put(55,29){\vector(-4,-1){14}}
\put(55,35){\vector(-4,1){14}}
\put(65,35){\vector(4,1){14}}
\end{picture} }
\end{picture}
\normalsize
\caption{HOL Sliding Window Protocol Model}
\label{fig:holstruct}
\end{figure}
\normalsize
To describe this model in HOL we must first define some types. Enter the
following definitions in HOL :
\begin{session}
\begin{verbatim}
let time = ":num"
and data = ":*"
and sequence = ":num"
and non_packet = ":one" ;;
let packet = ":(^sequence # ^data) + ^non_packet" ;;
let channel = ":^time -> ^packet" ;;
let seqtime = ":^time->^sequence" ;;
let datatime = ":^time->^data list" ;;
\end{verbatim}
\end{session}
The packet type is made up of either a sequence number, data message
pair, or a non-packet which represents a message lost or garbled by
the channel or a time at which no message is being transmitted.
The following
operators are used for creating, reading and writing variables of the
packet type.
\begin{session}
\begin{verbatim}
let set_non_packet = new_definition(
`set_non_packet`,
"set_non_packet:^packet = INR(one)");;
let good_packet = new_definition(
`good_packet`,
"good_packet (p:^packet) = (ISL p)");;
let new_packet = new_definition(
`new_packet`,
"(new_packet (ss:^sequence) (dd:^data)):^packet = (INL(ss,dd))");;
let label = new_definition(
`label`,
"label (p:^packet) = FST(OUTL p)");;
let message = new_definition(
`message`,
"message (p:^packet) = SND(OUTL p)");;
\end{verbatim}
\end{session}
The \verb+INIT+ predicate describes the initial values of the main program
variables and a condition for the range of sequence
numbers, \verb+maxseq+.
\begin{session}
\begin{verbatim}
let INIT = new_definition(
`INIT`,
"INIT (source:^data list) (maxseq:^sequence)
(rem :^datatime) (s:^seqtime)
(sink:^datatime) (r:^seqtime) =
( ( 1 < maxseq ) /\
(rem 0 = source) /\ (s 0 = 0) /\
(sink 0 = NIL) /\ (r 0 = 0))");;
\end{verbatim}
\end{session}
A channel is modelled by two streams, one for packets input to the channel
and one for those output by the channel over time.
The relationship between the two streams at any instant of time is that
the
packet on the output stream is either a copy of the packet on the input
stream or is a non-packet. The assumption that packets can be output
in the same time granule in which they are input is unrealistic
and eventually a channel model including message delivery delay will be
used instead.
\begin{session}
\begin{verbatim}
let CHANNEL = new_definition(
`CHANNEL`,
"CHANNEL (inC:^channel) (outC:^channel) =
(!t:^time. (outC t = inC t) \/ (outC t = set_non_packet))" );;
\end{verbatim}
\end{session}
The predicate \verb+DATA_TRANS+ represents the data transmission parts of
the SENDER program. Any packet from the
SENDER window is chosen for transmission.
The sender's window consists of the first \verb+SW+ elements of
\verb+(rem t)+.
The message chosen is the i-th element of \verb+rem+,
\verb+HDI (i t) (rem t)+, for \verb+(i t) < SW+. Its sequence number
is
\bigskip
\begin{center}
\verb+plusm(s t, i t, maxseq)+
\end{center}
\bigskip
\noindent where \verb+s t+ is the true sequence number of the first element
of \verb+rem t+. The functions \verb%plusm% and \verb%subm% perform
addition and subtraction modulo \verb%maxseq%.
The choice function, \verb+p+, guarantees
that \verb+i t+ is within the window and that there are at least
\verb+i t+
data messages available in \verb+rem t+. If \verb+p t+ is false then a
non-packet is transmitted instead.
With the intention of making this implementation model as general as
possible, only the minimum information needed about \verb+p+, its type,
is given at this stage. Later,
different implementations can be defined by more fully defining such
parameters.
The conjunct \verb+ ~(NULL rem t)+ is included in \verb+DATA_TRANS+ so
that, after all data has been transmitted,
non-packets are transmitted forever. That is, the
program never exits, but the protocol will reach a stable
state once transmission of source data has been completed.
\begin{session}
\begin{verbatim}
let DATA_TRANS = new_definition(
`DATA_TRANS`,
"DATA_TRANS
(rem:^datatime) (s:^seqtime) (SW:^sequence)
(maxseq:^sequence)
(p:^time->^sequence->bool) (i:^seqtime)
(dataS:^channel) =
( (SW = (maxseq-1)) /\
(!t:^time.
( ((p t (i t)) ==>
((~NULL(TLI (i t) (rem t))) /\ ((i t) < SW) )) /\
( ((p t (i t)) /\ ~(NULL(rem t)))
=> (dataS t = new_packet
(plusm(s t,i t,maxseq))
(HDI (i t) (rem t)) )
| (dataS t = set_non_packet ) ) ))) ");;
\end{verbatim}
\end{session}
For any packet, \verb+p+, \verb+(IN_WINDOW p b ws maxseq)+ tests whether
\verb+p+ is within the sender or receiver's window by testing whether
\verb+label p+ is less than window-size, \verb+ws+, sequence numbers from the
bottom edge of the window, \verb+b+.
\begin{session}
\begin{verbatim}
let IN_WINDOW = new_definition(
`IN_WINDOW`,
"IN_WINDOW (p:^packet) (b:^sequence) (ws:^sequence) (maxseq:^sequence) =
( (good_packet p) /\
( subm(label p,b,maxseq) < ws ) )");;
\end{verbatim}
\end{session}
The predicate \verb+DATA_RECV+ represents the part of RECEIVER program which
accepts incoming packets and outputs them to the sink if necessary.
This particular predicate uses a receiver window size of one which means
only packets which can be output immediately are accepted. If the incoming
packet, \verb+dataR t+ is acceptable, then its message part is appended to
the sink list, otherwise \verb+DATA_RECV+ makes no changes.
\begin{session}
\begin{verbatim}
let DATA_RECV = new_definition(
`DATA_RECV`,
"DATA_RECV (dataR:^channel)
(RW:^sequence) (maxseq:^sequence)
(sink:^datatime) (r:^seqtime) =
( (RW = 1) /\
(!t:^time.
(IN_WINDOW (dataR t) (r t) RW maxseq )
=> ( (r (t+1) = plusm(r t,1,maxseq)) /\
(sink (t+1) = (APPEND (sink t) [message(dataR t)] ) ))
| ( (r (t+1) = (r t)) /\
(sink (t+1) = (sink t)) )))" );;
\end{verbatim}
\end{session}
The RECEIVER program is also required to transmit information about its
current state to the SENDER in the form of acknowledgement packets.
The variables updated in \verb+DATA_RECV+ are used in the next acknowledgement
packet. Since \verb+r+ is the value of the next sequence
number {\em expected}, the
acknowledgement sequence is one less than \verb+r+ to represent the sequence
number of the last message correctly received. This strategy is chosen
so that the model generalizes for all window sizes from one
upwards. The boolean
function \verb+q+ is like \verb+p+ in \verb+DATA_TRANS+.
An acknowledgement packet is only transmitted at
times when \verb+q t+ is true.
The data field of an acknowledgement packet is never used, and can be
filled with any dummy data message.
\begin{session}
\begin{verbatim}
let ACK_TRANS = new_definition(
`ACK_TRANS`,
"ACK_TRANS (r:^seqtime)
(maxseq:^sequence)
(q:^time->bool)
(ackR:^channel) =
( !t:^time. !dummy:^data.
( (q t) => (ackR t = new_packet (subm((r t),1,maxseq)) dummy )
| (ackR t = set_non_packet) ) )");;
\end{verbatim}
\end{session}
Acknowledgement packets received by the SENDER program are checked by
\verb+ACK_RECV+. If the acknowledgement represents the correct reception
of a packet within the SENDER's window then a new window is chosen starting
from the first packet from the source stream which is as yet unacknowledged.
That is, the sequence number \verb+s t+ represents the sender's knowledge of
the latest message correctly delivered to the sink. The
first message {\em not} known to be correctly delivered has
sequence number \verb%s t%.
\begin{session}
\begin{verbatim}
let ACK_RECV = new_definition(
`ACK_RECV`,
"ACK_RECV (ackS:^channel)
(SW:^sequence)
(maxseq:^sequence)
(rem:^datatime)
(s:^seqtime) =
(!t:^time.
(IN_WINDOW (ackS t) (s t) SW maxseq)
=> ( (s (t+1) = plusm(label(ackS t),1,maxseq) ) /\
(rem (t+1) =
(TLI (subm (plusm(label(ackS t),1,maxseq),s t,maxseq))
(rem t)) ) )
| ( (s (t+1) = (s t)) /\
(rem (t+1) = (rem t)) ) )");;
\end{verbatim}
\end{session}
The \verb+SENDER+ predicate contains all those sections of the above
specification which would be part of the sender program in an implementation.
The conjunction of predicates represents their concurrent execution (this
is an abstraction of their real sequential execution).
\begin{session}
\begin{verbatim}
let SENDER = new_definition(
`SENDER`,
"SENDER (maxseq:^sequence)
(SW:^sequence)
(rem:^datatime)
(s:^seqtime)
(p:^time->^sequence->bool) (i:^seqtime)
(dataS:^channel)
(ackS:^channel) =
( (DATA_TRANS rem s SW maxseq p i dataS) /\
(ACK_RECV ackS SW maxseq rem s) )");;
\end{verbatim}
\end{session}
The liveness, or progress, of the sliding window protocol is monitored by
the \verb+ABORT+ predicate which would also be part of
the sender's program code in an implementation. This
predicate counts the lengths of times during which no
valid acknowledgement arrives. If any such waiting period exceeds
\verb+maxT+ then \verb+aborted+ is set to true and remains so.
Such a specification of liveness is called real-time liveness and
is based on the solution used in most implementations.
It is up to the program which is using
the protocol to decide what action to take when the \verb%aborted%
signal has been set.
The \verb+aborted+ variable is
used in the verification of liveness because
it can be proved that the protocol will eventually deliver all its data
as long as \verb+aborted+ remains false. Note that \verb+aborted+ will
be always false once all the \verb+source+ data, contained in \verb+rem+,
has been delivered. Throughout this specification termination is
represented by a predicate reaching and remaining in a stable state rather
than actually exiting a program.
\begin{session}
\begin{verbatim}
let ABORT = new_definition(
`ABORT`,
"ABORT (c:^time->num) (aborted:^time->bool) (maxT:num)
(maxseq:^sequence) (SW:^sequence)
(s:^seqtime) (rem:^datatime) (ackS:^channel) =
(!t:^time.
(c 0 = 0 ) /\
(c (t+1) = ( (IN_WINDOW (ackS t) (s t) SW maxseq) => 0 | ((c t)+1) )) /\
(aborted 0 = F) /\
(aborted(t+1) = (((c t >= maxT) \/ (aborted t)) /\ ~(NULL(rem t)))))");;
\end{verbatim}
\end{session}
The \verb+RECEIVER+ predicate, like \verb+SENDER+, represents all the
functions of the receiver program of an implementation.
\begin{session}
\begin{verbatim}
let RECEIVER = new_definition(
`RECEIVER`,
"RECEIVER (maxseq:^sequence)
(RW:^sequence)
(sink:^datatime)
(r:^seqtime)
(q:^time->bool)
(ackR:^channel)
(dataR:^channel) =
( (ACK_TRANS r maxseq q ackR) /\
(DATA_RECV dataR RW maxseq sink r) )");;
\end{verbatim}
\end{session}
Finally, all the implementation predicates described above are included in
a single predicate, \verb+IMPL+. A sliding window protocol
implementation is a system satisfying the behaviour of each of
its parts.
\begin{session}
\begin{verbatim}
let IMPL = new_definition(
`IMPL`,
"IMPL (source:^data list)
(maxseq:^sequence)
(rem:^datatime) (s:^seqtime) (SW:^sequence)
(p:^time->^sequence->bool) (i:^seqtime)
(c:^time->num) (aborted:^time->bool) (maxT:num)
(sink:^datatime) (r:^seqtime) (RW:^sequence)
(q:^time->bool)
(dataS:^channel) (dataR:^channel) (ackS:^channel) (ackR:^channel) =
( (INIT source maxseq rem s sink r) /\
(SENDER maxseq SW rem s p i dataS ackS) /\
(ABORT c aborted maxT maxseq SW s rem ackS) /\
(CHANNEL dataS dataR) /\
(RECEIVER maxseq RW sink r q ackR dataR) /\
(CHANNEL ackR ackS) )");;
\end{verbatim}
\end{session}
|