We are in Maseches Gittin, chapter 9, mishnah 4. The mishnah opens: "Sheloshah gittin pesulin" - there are three types of invalid gittin, and nevertheless "ve'im niseis - havalad kasher" - if she married on the strength of such a get, the child is valid. This needs explanation: Rabbi Meir holds that "kol hameshaneh mimatbe'a shetav'u chachamim begittin" - anyone who deviates from the fixed formula that the Sages established for a get - the rule is that he must divorce her and the child is a mamzer, meaning that as a rule she must leave the second marriage and the child born of it is a mamzer. Yet in these three cases even Rabbi Meir agrees that the child is valid, and their invalidity is only lechatchilah, from the outset.
The first get - "Kasav bichsav yado ve'ein alav eidim":
The husband wrote the get in his own handwriting, and there are no signatures of witnesses on it. Our mishnah follows the view of Rabbi Meir, who holds "eidei chasimah karsei" - the decisive element in a get is the actual signature of the witnesses on the document, unlike Rabbi Eliezer, cited at the end of the mishnah, who holds that the witnesses who observe the handing over of the get are the decisive ones. Even so, the husband's own handwriting is as good as a hundred witnesses, since the husband himself wrote it. Why then did the Sages invalidate it lechatchilah? For two reasons:
A decree: people might come to say that even a get written by a scribe who is not the husband and without witnesses is valid, whereas such a get is completely invalid.
It is like a get that has no date: since the husband writes it himself and there are no witnesses to the matter, he can write in it whatever date he wishes.
The second get - "Yesh alav eidim ve'ein bo zman":
The get has witnesses on it, but no date is written in it. Why does the time, that is, the date, need to be written in the get? Two views are given:
Because of his sister's daughter: as we mentioned previously, in a case where he was married to his sister's daughter and had compassion on her - even if she committed adultery, since she is his sister's daughter he would have pity on her and would postdate the get, so that the adultery would appear as though it occurred after the get was given, making it seem that she did not commit adultery, even though in truth she did. We therefore require a date in it, so that we know the correct date is indeed recorded, and he will have no ability to mislead.
Because of the produce of her property: if there were no date in it, the husband could continue eating the produce of her property even after the divorce, and we would not know exactly when the divorce took effect.
So we find that if there are no witnesses and the husband wrote the get, there is a problem; and if there are witnesses but no date on it, there is a problem in that as well.
The third get - "Yesh bo zman ve'ein bo ela eid echad":
The date is written in the get, but there is only one witness on it. There are two ways to understand the case:
It refers to a get written by the husband himself, and the novelty is that this is not only comparable to the first case, but even when there is one witness it is still invalid lechatchilah.
It refers to a get written by a scribe, and even though it was written by the scribe and has only one witness on it, if she married with it the child is valid, because we join the scribe as a second witness, and together they are as two.
And so the mishnah concludes: these are three invalid gittin, and one may not marry on the strength of them; however, "im niseis - havalad kasher" - if she married, the child is valid.
Why is the number of gittin counted twice?
The Gemara notes that the mishnah states the number of gittin twice - at its opening and at its close - even though the gittin are listed and detailed, and there is no need to repeat that these are the three gittin. The Gemara explains that both counts come to exclude something:
The count in the first line comes to exclude an old get, which we discussed in chapter 8: a woman who was divorced with a get after whose writing the couple continued living together, or at least were secluded together - such a get is invalid lechatchilah, but once she married with it she is permitted to remain married to the other man. It is therefore not included in our mishnah, which deals with women who cannot marry on the basis of such a get, whereas with an old get one may marry after the fact.
The count also comes to exclude the opposite extreme: a get written with the date of the wrong kingdom, which we also mentioned in chapter 8 - such a get is invalid even after the fact and the child is a mamzer, whereas our mishnah teaches that these gittin are different from it, in that the child is not a mamzer.
The count at the end of the mishnah comes to exclude a get that can be corrected, such as an agent who came from overseas and did not say "in my presence it was written and in my presence it was signed" - for he can take the get back and say this when he gives it to her a second time.
The view of Rabbi Eliezer:
Everything said until now follows the view of Rabbi Meir, who requires witnesses of signature. Rabbi Eliezer says: "Af beshe'ein alav eidim, ela shenasno lah bifnei eidim - kasher" - even where there are no witnesses on it, but he gave it to her in the presence of witnesses, it is valid. Even though no witnesses are signed in the body of the get, if he gave it to her in the presence of witnesses - these are the witnesses of delivery - it is a valid get. Moreover, "vegovah minechasim meshuba'im" - she can collect her kesubah with this get even from property that the husband sold to a third party, since her claim to that property precedes the sale.
And why are the witnesses who observe the delivery considered the essential element? Because "she'ein ha'eidim chosmin al haget ela mipnei tikkun ha'olam" - the signing that we generally practice, in which the witnesses write their signatures in the body of the get, is only because of tikkun ha'olam, out of concern that we will not have access to those witnesses who saw the delivery, since their actual presence is needed so that they can testify when the need arises. When the witnesses are signed on the get, the signatures remain with the get, unlike witnesses of delivery, who must be brought physically in order to be present.