An essay on Suggestivism
Suggestivism is not of course a real word. And that is rather the point of this essay. Suggestivism is an approach to cryptology in which one person’s authorship of a text or other document is influenced by suggestions as to the content of the text conveyed by the party with which one is communicating (or with which one imagines one is communicating). So for example, an email or word-processing document is written by one party and suggestions as to the appropriate text or ideas that should go into the document are suggested by the other party as the first party is writing the document. This does not work, and it descends into nonsense, which is why using the neologism suggestivism is so appropriate for this sort of attempt at cryptology.
Let us begin by giving an account of the archetypal kind of suggestivist technology. For it is an approach that works only where both parties to the communication have electronic devices with the capacity in principle to undertake electronic communications wirelessly and in conjunction with an external data source such as a WiFi connection, a mobile data signal, or potentially even something like a Bluetooth connection. Typically the principal author of the document, who we shall call the “suggestee” (the party to whom things are being suggested), is composing an email or WORD document. The “suggestor” (the party who is making remote suggestions into the author’s text.
The way this might be done is by highlighting pieces of the text in different ways (for example with a shadow), deleting words the suggestee types, providing word choices when the suggestee is partially or completely through the typing of a word, suddenly rearranging the windows on the screen of the suggestee’s device to highlight a particular word or sentence or group of words or to have the document disappear entirely, and so on and so forth. In particular, on certain sorts of mobile telephone a “word option” functionality in which three words are offered by the mobile phone when typing an email or instant message, for example; this functionality can be manipulated by rewriting the coding of the mobile telephone so that the suggestor can suggest three words instead of the mobile telephone software. This latter kind of suggestivism, that we will call the “three-word approach”, is vastly more open-ended and flexible than the laptop-based suggestivism that tends to offer to fill out incomplete words or offer changes to complete or incomplete words in various ways, for example with the word offerings in different shades of boxes or with different colours of underlinings of the words that when you click upon may make suggestive changes, and so on and so forth. Both methods also often involve intentional misspellings of words, and this adds both opportunities for discreet expression and opportunities for confusion, as we shall explore.
This method began a number of years ago with an approach in which the parties to a communication they wished to conceal would share the login details of a common email account and would add to and modify draft emails that were not sent. In that environment, the process of suggestion was rather easier because the suggestor could say exactly what he meant, working to the assumption that the communication was indeed confidential. The reason this relatively simple method of confidential communication broke down (and the simple methods are generally the best) was because it became possible, with technological advancement, to track who was opening specific email accounts and editing drafts; it even become possible to track the passwords of those email accounts, and therefore hostile parties could interfere with the process and this simple approach to use of email addresses to engage in confidential communications was gradually withdrawn in favour of something more sophisticated. And that development is what this article discusses.
We also observe that the method for the suggestor to make his or her suggestions is inevitably a mix of artificial intelligence and human input. In many cases the word choices come up so quickly that no suggestor typing on a keyboard could deliver them so quickly. Nevertheless the word choices offered - and this is particularly so with the “three-word approach” - reflect specific comments, observations, assertions warnings, instructions et al that the suggestor wishes to convey. Therefore the suggestor enters these words, phrases, concepts or sentences into a piece of software, that does its best to convey the meaning the suggestor wishes to convey in a manner at least ostensibly consistent with the operation of word prediction software, so as to preserve plausible deniability at least as far as is possible consistently with the urgency of the communication. There is also a manual override in cases of grave emergency, for example when the suggestor sees the suggestee entering text that the suggestor really does not want the suggestee to enter for whatever reason.
The AI-supported approach is all the more essential when using the three-word approach because the suggestee can be entering text so extremely quickly ((s)he may just be tapping on the word suggestions at great speed, rendering rapid meaningful manual interaction impossible). By contrast on a laptop or other device without the three-word approach capacity, the speed of typing may be slower and there is a far greater opportunity for manual interaction. Nevertheless the speed with which a person types may be significant, so the suggestor’s software may need to be constantly suggesting amendments for the suggestor to approve manually (perhaps from a list of pre-inputted words concepts / phrases / sentences) and/or the suggestor may have to rely entirely upon artificial intelligence to convey as best as is possible suggestions that indicate to the suggestee the direction in which his or her authorship should be going.
Finally, it is desirable (even if this is not typically implemented) that the suggestor’s software interface have an indicator for degree of familiarity, experience or expertise with any particular suggestee’s use of a suggestivist method of communication, whether set manually or via an AI-based computational method. On one possible analysis, but there may be others, the more sophisticated and competent the user of the suggestivist method in question, the more subtle the suggestions need be and hence plausible deniability is more adequately maintained in the suggestions made; and vice versa.
With these opening observations about suggestivism now in place, we proceed to consider a series of problems with the approach that, taken in toto, suggest why a nonsensical neologism is an appropriate term to describe the suggestivist method of encrypted communication. These observations are made in no particular order.
The agenda issue
We begin by citing the agenda issue. The suggestee must have an agenda to work to or he does not know what to write about (or, better, to start writing about) because otherwise (s)he will be at a loss as to expect what suggestions the suggestor might make and the whole process will begin in a bundle of confusion and potential mutual miscommunication, in particular where the three-word approach is being used. All sorts of strange words might start being suggested, and without a pre-set agenda the suggestee does not know the context of these suggestions and can barely be expected to have any idea what they mean. For the “laptop approach”, one can barely get started at all without an agenda as the suggestee simply has a blank sheet of paper with which to begin his authorship.
How does one establish a pre-set agenda? Not by suggestivism as this would commit the fallacy of petitio principii. In other words, the effort is circular.
Hence the only ways of establishing an agenda are (a) the suggestee unilaterally establishes an agenda, ideally in advance in writing something that the suggestor can read; or (b) at the end of a prior conversation. The problem with that is how the parties establish that they have finished discussing the issues of substance from the current discussion and are now moving onto the agenda for the next discussion. The suggestee may do this definitively, by saying “now I want to talk about the agenda for the next discussion”; the suggestor may do it on the three-word approach if they are a sufficiently sophisticated user but there is ample scope for confusion and miscommunication, for example if the suggestee thinks that actually the suggestor wants to add some point (s)he missed out of the primary discussion. All in all, this problem is an extremely hard one, and it tends towards unidirectionalism: that is to say, a form of communication in which only the suggestee can communicate towards the suggestor because no agenda can reliably be set which gives sufficient context to the meaning of the words for the suggestee to communicable reliably to the suggestor. And as soon as we reach unidirectionalism, there is no point in suggestivism at all. So the agenda problem is critical to resolve.
The timing issue
When will the suggestor and suggestee communicate? At what times are they due to communicate? One option is to have the parties agree a fixed time each day (for example) when they will communicate: either in advance of the regular communications blackout that gives rise to the need for cryptological methods; or subsequently by habit, explicit statement or otherwise.
The principal problem with the “fixed times” approach is that it increases the risk of the “adversarial or nuisance input” issue, which we will discuss further below. For now let us just observe that because the various suggestivist methods appear to be transparent, in the sense that other third parties, if aware that suggestivist communications are taking place, may intervene in them without the suggestee at least and possibly even the suggestor necessarily knowing, the risks of this taking place increase exponentially if there is a fixed schedule for communications. That is because the interfering party, by one means or another (either electronic observation or traditional human intelligence techniques - i.e. watching somebody physically), may learn of the routine and this knowledge will render it far easier to interfere effectively.
Hence, particularly where the suggestee has a high profile within communities interested in the sorts of matter discussed in this essay, and hence the possibility of actively looking for opportunities to interfere is all the higher, the “fixed time” approach will not do.
Hence once may be forced to resort to use of arbitrary times where one suspects or conclusively reasons that there is interference, in particular at strange times such as in the middle of the night when the suggestee imagines that the intervenor may not be awake. However this gives rise to the need for 24-7 monitoring by the suggestor or his/her colleagues, and this in turn needs a team of at the very least three to four (because the process of suggesting requires concentration, skill, and is time consuming and mentally exhausting, particularly where tricky or complex messages are intended to be conveyed) plus constant updating of the suggestee’s dossier to understand what is on his mind and this in itself may not be able to be an exclusively computational exercise because computers are not always good at establishing a person’s priorities or even their identity (for example in the case of stolen or compromised suggestee electronic equipment). So the whole thing ends up taking a team of minimum five people working full time, seven days a week. It is therefore highly consuming of resources - in in plainer terms, very expensive.
An alternative is to switch the suggestor function onto 100% AI functionality at certain times of the day / week etcetera, but this is very dangerous in case some urgent or critical communication starts to be received from the suggestee and the AI algorithm does not spot it and does not alert a relevant human operator. Also people get hacked off when they realise they are just speaking to machines (and most humans of reasonably high intellect can still tell the difference between an electronic interaction with a genuine human element and one that is simply a conversation with a bot), and then they will cease to use the suggestive tool and the system will fall into disuse.
Even an intervenor may pick up arbitrary times, by writing and/or deploying computer software that detects specific types of activity on the suggestee’s electronic device. So one experience it is possible to suffer is to realise that an aggressive intervenor is alerted to intervene after so much time period of a certain sort of electronic activity on a suggestee’s device has elapsed. Hence the suggestee writes a very short / quick message, stops, does something else, and then writes another short / quick message. The problem with this is that it tends again towards unidirectionalism, because the less time you think you have to write your point before interference begins, the less time and concentration you will give to any suggestions coming from the suggestor.
The principal problem with suggestivism is and always has been ambiguity, particularly where messages are even remotely complex and hence where accuracy of the written word is essential in establishing what the suggestor means. Because the greater majority of suggestions can be ambiguous, in both the three-word approach and the laptop approach.
We pick a simple example. Suppose the suggestee is in the middle of writing the word “Summer”, and after typing the second “m” (s)he is faced with the word choice “Sung”. This is obviously not a standard word correction error or suggestion; it is obviously a piece of suggestivism. But what does it suggest? “Sung” could (just by way of example; the following is this author’s own list) any of (a) sex noises (high-pitched singing); (b) sunglasses (“Sung” being the first letters of another word); (c) hot weather (same logic as (b)): (c) looks cool (whatever that means) (same logic as (b)); (d) someone has confessed; (e) someone has betrayed somebody else by giving away information that renders the suggestee (or someone else) at risk; (e) “Sung his praises”, i.e. the person is an effective agent; (f) given up (again a reference to high-pitched noises). The whole exercise of guessing at what is being suggested descends into nonsense.
The ambiguities problem is both diminished and enhanced if the three-word approach is adopted. It is diminished because the suggestor can proffer three words that read together may reduce the possible ambiguity. It is enhanced because if the suggestee does not see the connection between the three words then (s)he may construe the three words as meaning something different: for example giving a choice to the suggestee or granting the suggestee an opportunity to express his preference about something. Moreover the suggestee may be confused as to whether a suggested word ought to be clicked on in order to receive more words to continue making sense of whatever it is that the suggestor is suggesting; or whether it reveals something but should not be clicked on in order to preserve confidentiality and instead some other word should be clicked on in order to move onto a new subject or to reveal more information about the current subject.
The gravest ambiguity in the three-word approach is that it is not clear to the suggestee at what points (s)he ought to stop clicking upon the various words offered and instead start offering his own words and seeing what happens.
Then there is the problem of ambiguity in relation to what we might call “buffers”. These are three words in the three word approach that collectively make no sense, for example “Police police Russians”. This suggests something. But it might suggest all sorts of things (aside from the fat that the Police are Russians or that the Police know you are Russian etcetera etcetera). It might mean that this is the end of this subject; this is the end of this conversation; you have clicked on the wrong words and achieved a miscomprehension and should go back; and so on and so forth. So even buffers turn out to be highly problematic.
As a general principle, when faced with ambiguities people tend to adopt an interpretation of the ambiguous word, phrase or indication either highly agreeable to them (i.e. an outcome they want) or one that represents something they fear. The suggester cannot predict which of these two directions the suggestee is going to fall towards. It may depend upon his or her mood, which the suggestor can seldom constantly and accurately monitor even with the most advanced contemporary profiling techniques and mood identifiers based upon artificial intelligence techniques burying through things like CCTV and microphone data. It is an art, not a science, to assess someone’s mood.
In conclusion, the ambiguities problem is so agonising that it renders suggestivist cryptology excruciatingly difficult and requiring massive amounts of concentration and intellect; and even then it can be confusing, demoralising, exhausting and frustrating. And that is before we get to a series of specific issues related to the ambiguities problem, that we shall consider in turn.
The referential issue
This is a species of the ambiguity issue, that deserves brief separate mention. The point is that if an object adjective quality or property is referred to (“the referred item”), then it is inherently ambiguous as to to whom which or what the referred item is related to absent unambiguous use of possessive pronouns or punctuation marks indicating ownership.
We shall give two examples. One is “auto”. Now let us assume that the suggestor’s use of “auto” is a reference to a vehicle - even, perhaps, what in British English is known as a car. (It could be a reference to various other things - for example an automated computer process; an automatic gearbox in a car; a motorway or highway (“Autoroute” in German and French) - and so on). Whose car is it? Is it the suggestee’s car? What if (s)he doesn’t have one? Should (s)he go and buy one? For what purpose? Does it mean a car is being made available? Where, and for what purpose? To go to the airport? To execute some driving in the style of a bad movie? Or is it someone else’s car? If so then whose? And why is it being referred to? Should the suggestee go and get in this car? Should they not get in it?
Then we get onto the problem of possessive pronouns, that might be used to clarify the referential issue. This can be a mixture of nightmares all of its own. What does “your” mean? Does it mean (a) the suggestee; (b) the suggester (after all the word is coming up on the suggestee’s electronic device, so it may be construed as a reference to the person with whom the suggestee is communicating): (c) the suggestee’s team (whatever that might mean); (d) the suggestor’s team? And so on and so forth. The only real way of solving the possessive pronouns issue is not to use possessive pronouns at all but instead to use people’s actual names or comprehensible abbreviations or euphemisms for them. But that risks giving away the names of actual people as a result of electronic communications interference of any kind, which is becoming ever more prevalent these days as we shall discuss in brief below.
The second example of the referential issue ambiguity we proffer is the word “blue”. So the first problem is that the word “blue” might not be a reference to a colour at all but instead to (a) sexual activities (a blue movie); (b) being depressed or in poor mental health (he had the blues); (c) a sort of musical style; (d) a body of water; (e) the sky (and possibly aviation?); and so on. But the more fundamental problem is that if “blue” is a reference to a colour then you need to make it clear what sort of thing it is the colour of. Moreover that item had better be in the suggestee’s immediate line of sight, or (s)he is not going to have the slightest idea what to make of the colour reference. After all, lots of things one comes across in life are blue, and you can’t spend an indefinite period of time wondering around looking for blue things just because that’s what the suggestor might have meant.
The spelling mistake issue
The spelling mistake issue is a fairly obvious one, namely that a typist (particularly a quick one) may be unable to distinguish between a mobile telephone’s or laptop’s genuine attempt to correct erroneous spelling or grammar; and a piece of suggestivism on the part of the suggestor. Hence when the typist sees a suggestion flash up, then often disappear again because (s)he may be typing fairly quickly, (s)he has to apply backspaces to the text and see whether the suggested text appears again, and if so then what consequence to draw from that. Is the suggestor merely suggesting in flashing up a word with the correct spelling that the suggestee is heading in the right direction with the text (s)he is writing? There are and can be no correct answers to these questions.
After all, as soon as we wrote a private dictionary about the issue someone would sell it to our opponents or they would otherwise work out what it was through a computational monitoring process, and hence we revert to the problem that all synchronous cryptology - that is to say, cryptology with unambiguous private dictionaries - can be decrypted using advanced computational methods.
The spelling mistake issue is more of a problem on laptop suggestivism than three-word suggestivism on mobile telephones, because the only method of the suggestor rendering the suggesting is through imagined spelling mistakes in what is actually being typed (plus a few more crude tools such as flashing blocks of text but it is typically extremely difficult to understand what things like that mean), whereas the three-word approach is less reliant upon correcting spelling errors.
The non-functional communications issue
Sometimes the suggestions are intended as light-hearted observations upon the author’s style, personal remarks between the communicating parties with no relevance to the professional task at hand, or even commendations or criticisms of what the suggestee (i.e, the principal author) is doing.
While all these things are often welcomely received by the author, especially if (s)he is operating in a lonely environment, the capacity to use suggestivism less unambiguously is diminished if this sort of usage is permitted (and it is difficult to see how it could be prohibited). That is because the suggestee may easily be confused as to whether the nature of the communication rendered by the suggestor is pertinent to the text (s)he is writing; pertinent to the propriety of the document as a whole (e.g. is the suggestor suggesting that the document be deleted in its entirety?); or just a welcome casual remark. Or indeed it might be construed as an unwelcome casual remark from an adversarial intervening suggestor. Also a casually amusing remark or any other kind of comment might be interpreted offensively by the suggestee, diminishing his morale in a lonely environment.
Hence great care must be taken with any non-functional communications rendered from the suggestor to the suggestee, although such comments are typically heartily welcomed. The most manifestly unpleasant ones are typically considered by a moderately competent suggestee to be associated with hostile interference in the communications, that we consider below. So as a general rule, because people like to associate ambiguity with something good, pleasant or warm contributions
Meetings, times and places issue
Both forms of suggestivism considered in this essay have tremendous trouble identifying that a meeting should take place and if so where, or referring to times and places generally. There is a general reluctance in the suggestor software coding to permit the suggestor from directly suggesting a time or a location. And without a time and location, one cannot of course arrange a meeting, which people in field locations may often want to do for a variety of reasons.
The reason surely why suggestor software does not let suggesters as a rule directly suggest times or places is due to the risk of those phrases being overheard by virtue of electronic communications interference, that we will discuss in the next section of this essay. Due to the ease of that interference, the risk is that a hostile party might intercept straightforwardly the time and place for such a meeting, and then potentially dangerous or otherwise hostile parties might intercept the meeting, either at its appointed time, or beforehand, or afterwards.
Nevertheless meetings do have to be arranged, and if there is ambiguity in the language used to arrange them then it is highly likely that the meeting will not take place as planned because the suggestor and the suggestee drew different inferences from an inherently ambiguous exchange. Nor can this problem be resolved by a synchronous private dictionary, for the same reasons as previously discussed with such materials: they can either be bought or learned.
The conclusion we draw from this brief discussion is that unless discussed in a secure environment (for example in advance of mission), it is not possible to arrange meetings, times or places for any activities using electronic means other than by unencrypted communications. And you have to take the risks associated with the possibility that a hostile party might intercept those unencrypted communications, and you have to mitigate them. Or you have to use a non-electronic means of communication to convey such details, for example an analogue telephone call (if you can have reasonable confidence that the call is not being tapped, an increasingly infrequent skill in the modern electronic world) or just something old-fashioned like a piece of paper written in distinctive and unforgeable handwriting. (Handwriting forgery is likewise an increasingly rare skill in this modern world dominated by electronic communications.)
What is to stop an adverse third party from participating in the suggestivism process? The answer is regrettably little, and this is a monumental problem with electronic suggestivism of every kind. It is not a hard problem to state, but it is an exceptionally hard problem to solve and so far nobody appears to have been able to solve it although there are some pioneering efforts underway that are outside the scope of this essay.
The problem is this. Suggestivism works by exploiting holes in the electronic communications software and firmware of contemporary electronic devices, and rewriting the code in computer programmes such as WORD or email software to change the direction of suggestive wording. Now if one party, standing behind the suggestor and suggestee, can do this, then another party, whose interests are averse to those of the party standing behind the aforementioned parties, can also do it and start engaging in all sorts of hostile, offensive, threatening, fraudulent, fake and other dangerous suggestions. This may result in confusion and disorientation. It may become worse; it may cause mental health problems, in particular anxiety, as the suggestee fails to understand (a) that (s)he is now communicating with an adverse party, and hence takes damaging messages in good faith; or (b) does understand that from time to time (s)he is now receiving communications from an adverse party, but cannot tell which one. In extreme cases, this may result in psychosis. The process of flipping from confusing messages from a friendly suggestor to confusing messages to an unfriendly suggestor, and back again - may drive a person to start believing things that are not true. (This is a simplified definition of the psychiatric condition of psychosis: belief in falsehoods for which there is no proper evidence.) When one’s source of information is something so confusing and potentially malicious, then it is not difficult to flip into psychosis.
Another extreme danger is that an adverse interferer in the suggestive process suggests a course of action to the suggestee that is dangerous or harmful, and leads the suggestee to place his life or welfare in danger.
The party most prone to this type of electronic communication interference, with such deleterious consequences as causing mental health issues, chaos and confusion with mission projects, and potentially grievous bodily harm or even death, is the various intelligence agencies of the Russian Federation. And it is a constant battle against them, to prevent their toxic attempts to interfere with confidential communications from causing additional expense, inconvenience, chaos, danger or injury.
At the current stage of technological development it appears that suggestors have established the capacity to tell when a hostile intervenor is intervening in the suggestor process; but they cannot prevent it. At best, a “queue” of differential suggestive messages is formed, some sent by benign suggestors and others sent by malevolent suggestors; and they are delivered to the suggestee all mingled up. This is enormously confusing, distressing and dangerous for the suggestee, as already noted. Efforts seem to be underway to assist the suggestee in differentiating between benign and malevolent suggestions, but it appears to be in a prototype version, involving a lateral indication when there is a danger of interception or actual interception. Nevertheless the suggestee in deploying this approach finds themselves constantly shutting down their electronic device, which disrupts thought flows and also risks the return to unidirectionalism, discussed above, because each short message burst is interrupted and there is insufficient time for the suggestee to consider the possible meaning and context of the various legitimate suggestions being made.
This leads into an identity crisis more generally, in which different benevolent (or assumed benevolent) actors may simultaneously or sequentially engage a suggestee; their messages may become mingled; there is no clear method for establishing when one benign suggestor signs off and another one is signing on; how these multiple benign suggestions are interacting with potentially multiple malign suggestions; nor indeed who a suggestee is communicating with at all at any particular moment.
What is really needed is two things, for such a system to work at all. One is an identity security protocol. You must know who it is you are communicating with, and this must be ascertained with absolute certainty and not just by reference to speech patterns or something similar because those can be copied and anyway the suggestor may be someone you have never met or spoken with before and hence the whole exercise is just a shot in the dark. Absolute clarity of the identity of the communicator must be achieved even to attempt to use the suggestivist messaging system, because so much ambiguity in meaning derives from the context pertinent to the identities of the individuals communicating and the situations they are in or may be imagined to be in. So for example, the word “blue” might indicate a blue movie to some people but might well not do so to a range of others. It all depends upon the individual. And actually every word in every language is more or less like this. It is a quality of language that the American philosopher Willard Van Orman Quine observed in his theory he called “the indeterminacy of translation”. In his classic example, the sentence “it is raining” is not necessarily true if and only if it is raining, because the words “it”, “is” and “raining” can all mean different things depending on the context of the communication and the identity and knowledge of the communicating parties.
The second tool such a system of communication needs is an absolute block-out for other intervenors while two communicants engage. Otherwise there is always a risk of miscommunications because new parties with new identities are poking themselves into an ongoing discussion and this will cause the suggestee to become confused and lose confidence in suggestivism as a method of communication and then we are back to unidirectionalism or indeed no cryptology at all.
Hence suggestivism using electronic means leads to a crisis of identity - we do not know who we are and hence we do not know what we are talking about and hence all social rules and social structure break down. For if the suggestor’s communications can be interfered with, then so of course can the suggestee’s, and fake emails can be sent in the suggestee’s name or in the names of other people and all sorts of other strange things can happen that will then thoroughly confuse everybody. At this point, the very act of communication becomes worthless; there is no sense in anything you write, because it cannot verifiably come from you and nobody can understand what it really means. At this point, only actions make any sense and words and language become useless. This is of course a bad thing, because ultimately the only actions that make any sense are making love, protecting people from violence and inflicting violence on others. And that is the law of the jungle; it is what distinguishes we humans from the animals.
Another way of looking it the problem is through the lens of what the German sociologist Niklas Luhmann called a “legitimation crisis”. In the absence of adequate and clear communications in a bureaucracy or in any other form of social organisation that requires more than one person, the lack of clear and unambiguous communication renders the legitimacy of the institution at issue in peril. Language must be tied down and everything must be as clear and unambiguous as possible. The identity of each person must be clear and it must be transparent who is responsible for what. Sensible, clear, plain rules must be drafted and followed, including plain clear speech; or institutions cannot function because they become lost in a haze of ambiguity and crypticism. Reading Luhmann, whose work is notoriously dense and impenetrable, one might be forgiven for concluding that he had not bothered to learn his own lesson. Nevertheless his work provides valuable insights in the fields of cryptology and communication. His point is that clear, unambiguous, plain words are essential to avoid a legitimation crisis, even if indeed it does mean that everybody can understand them. That is, of course, the very point of language.
At the current juncture, it is very tempting to suggest that we all just go back to using pens and paper. Security in electronic communications at the current time is highly problematic.